Release gardening

- add `lint` target to tox as suggested by @bertjwregeer to verify that manifest is ok,
- add `RELEASING.txt` to document how to make a new release

/cc @jamadden, @azmeuk, @stevepiercy 
/reviewed-by @stevepiercy 
/reviewed-on https://github.com/Pylons/zodburi/pull/30

@stevepiercy says (https://github.com/Pylons/zodburi/pull/30#discussion_r630521139):

    This should be a plain text file, no markup. The intended audience are
    maintainers. It should not get included in user documentation. Plain
    text makes it easy to maintain and open links in editors that support
    that feature.
    ...

-> Rework the text based on that suggestion.
-> Keep pointers on how to do it manually for the reference.

Based on suggestions from @stevepiercy:

- split lint into lint (doing only checks) + build - to build
  sdist/wheel to distribute:

  https://github.com/Pylons/zodburi/pull/30#discussion_r630537231

- structure list of environment for readability:

  https://github.com/Pylons/zodburi/pull/30#discussion_r630523259

Add minimal document that describes how to make next zodburi release.
As suggested by @jamadden zest.releaser, by default, works ok most of the time.

Add minimal linting step that verifies manifest and built sdist/wheel.
Suggested by @bertjwregeer on https://github.com/Pylons/zodburi/pull/29#issuecomment-836915440.

While trying to prepare next zodburi release I've noted that the source tarball
- generated by `python setup.py sdist` or by `python -m build` - is missing some
files compared to source tarball of previous zodburi release uploaded to PyPI:

https://pypi.org/project/zodburi/2.4.0/#files
( https://files.pythonhosted.org/packages/a2/dc/ab3f339d6e123417d7f259a79395a8f6793478d5089886a8b00d186293b1/zodburi-2.4.0.tar.gz )

Many files like CHANGES.rst, contributing.md, CONTRIBUTORS.txt, tox.ini
and so on were missing:

    $ xdiff --stat x/zodburi-2.4.0 zodburi-2.4.1.dev0
    x/zodburi-2.4.0/.gitignore => /dev/null                                   |  14 --
    x/zodburi-2.4.0/.travis.yml => /dev/null                                  |  35 -----
    x/zodburi-2.4.0/CHANGES.rst => /dev/null                                  |  98 ------------
    x/zodburi-2.4.0/CONTRIBUTORS.txt => /dev/null                             | 105 -------------
    x/zodburi-2.4.0/COPYRIGHT.txt => /dev/null                                |   2 -
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/PKG-INFO                          |  17 +-
    x/zodburi-2.4.0/contributing.md => /dev/null                              |  23 ---
    x/zodburi-2.4.0/docs/Makefile => /dev/null                                |  85 ----------
    x/zodburi-2.4.0/docs/api.rst => /dev/null                                 |  10 --
    x/zodburi-2.4.0/docs/conf.py => /dev/null                                 | 205 ------------------------
    x/zodburi-2.4.0/docs/index.rst => /dev/null                               | 385 ----------------------------------------------
    x/zodburi-2.4.0/rtd.txt => /dev/null                                      |   1 -
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/setup.py                          |   7 +-
    x/zodburi-2.4.0/tox.ini => /dev/null                                      |  35 -----
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi/__init__.py               |   7 +-
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi/resolvers.py              |  45 ++++++
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi/tests/test_resolvers.py   | 147 ++++++++++++++----
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi.egg-info/PKG-INFO         |  17 +-
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi.egg-info/SOURCES.txt      |  12 --
    {x/zodburi-2.4.0 => zodburi-2.4.1.dev0}/zodburi.egg-info/entry_points.txt |   1 +
    20 files changed, 203 insertions(+), 1048 deletions(-)

-> Fix it by explicitly listing which files need to be distributed in MANIFEST.in

The MANIFEST.in is based on MANIFEST.in from waitress:

https://github.com/Pylons/waitress/blob/master/MANIFEST.in

Not sure how previous source tarballs were prepared - probably it was some kind
of per-user global MANIFEST being taken on Tres' development machine previously.
They say explicit is better than implicit anyway...

P.S. I did not included .travis.yml and .gitignore into the MANIFEST because
those files should not be distributed in tarballs.

/reviewed-by @bertjwregeer
/reviewed-on https://github.com/Pylons/zodburi/pull/29#issuecomment-836915440

- Both ZODB4/ZODB5 were started to be supported since 7502fa4f (tox tests with zodb4),
- PyPy started to by supported since 6c46a29d (Added pypy to tox and travis),
- ``demo:`` URI resolver was added in 2f02bc90 (demo: URI resolver).

Bump next version due to the new features.

/reviewed-by @bertjwregeer
/reviewed-on https://github.com/Pylons/zodburi/pull/29#issuecomment-836915440

demo: URI resolver

This patch adds `demo:` URI scheme to create DemoStorage from an URI.

While several existing resolvers already handle ?demostorage argument to
wrap itself with DemoStorage and in-RAM MappingStorage for changes, this
approach has the following drawbacks:

- every resolver must do it
- it is not possible to create DemoStorage with non-MappingStorage for changes.

My particular motivation here is Wendelin.core 2: it spawn WCFS
filesystem server to serve array data from ZODB storage, and passes
storage URL to spawned wcfs process, so that wcfs could connect and
retrieve data from the same ZODB storage that client process is
using. When original ERP5 client is using DemoStorage, both `base` and
`changes` must be persisted because if changes would be in-RAM
MappingStorage, WCFS could not access that data because it runs as a
separate process.

To build a DemoStorage URI we follow XRI Cross-references approach to
embed URIs for base and changes into combining demo URI:

    demo:(base_uri)/(δ_uri)

https://en.wikipedia.org/wiki/Extensible_Resource_Identifier provides
some related details and examples.

I choose fragments as the place for ZODB.DB arguments:

    demo:(base_uri)/(δ_uri)#dbkw...

The reason fragments - instead of parameters - are used, is because
DB arguments are _local_. Even with different DB arguments the URI
refers to the same storage, and for wendelin.core 2 it is important to
be able to determine whether two client processes actually use the same
underlying ZODB storage, even if those two clients use different local
parameters, like connection_pool_size and similar. Keeping such local
arguments in fragments makes it easy to determine the underlying URI of
the storage - by dropping fragments. It also follows logic from RFC 3986,
which says that "fragment ... may be ... some view on representations of
the primary resource": https://tools.ietf.org/html/rfc3986#section-3.5

Thanks beforehand,
Kirill

/cc @tseaver, @azmeuk, @jimfulton

P.S.

It was already suggested several times to add DemoStorageURIResolver and
deprecate ?demostorage in existing URI resolvers:

https://github.com/Pylons/zodburi/pull/25#issuecomment-485506959
https://github.com/Pylons/zodburi/pull/25#issuecomment-511480572
https://github.com/Pylons/zodburi/pull/25#issuecomment-622931793

- Drop support for 3.4.
- Add support for 3.8

Co-Authored-By: Tres Seaver <tseaver@palladion.com>

Added pypy and pypy3 tests

Add tests for zodb4.

Garden changelog since 2.3.1.

Annual housecleaning

Fix PendingDeprecationWarning about cgi.parse_qsl

- Do not use smart quotes

- Perpetual year in copyright
- Current version
- Clean up imports and calls

I was running py.test based tests on a project which uses zodburi and
got many warnings:

	.../python2.7/site-packages/zodburi/resolvers.py:86: PendingDeprecationWarning: cgi.parse_qsl is deprecated, use urlparse.parse_qsl instead
	  kw = dict(parse_qsl(query))

Since for Python2 cig.parse_qsl was migrated to urlparse.parse_qsl in 2008:

	https://github.com/python/cpython/commit/c585df9476

it is safe to change to urlparse.parse_qsl unconditionally.

Fix typo in Trove classifiers.

zeo: Fix parsing URI with IPv6 address

The code was crashing this way:

---- 8< ----
======================================================================
ERROR: test_call_ipv6 (zodburi.tests.test_resolvers.TestClientStorageURIResolver)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/kirr/src/wendelin/venv/z-dev/local/lib/python2.7/site-packages/mock/mock.py", line 1305, in patched
    return func(*args, **keywargs)
  File "/home/kirr/src/wendelin/z/zodburi/zodburi/tests/test_resolvers.py", line 265, in test_call_ipv6
    factory, dbkw = resolver('zeo://[::1]:9090?debug=true')
  File "/home/kirr/src/wendelin/z/zodburi/zodburi/resolvers.py", line 141, in __call__
    host, port = netloc.split(':')
ValueError: too many values to unpack

======================================================================
ERROR: test_call_ipv6_no_port (zodburi.tests.test_resolvers.TestClientStorageURIResolver)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/kirr/src/wendelin/venv/z-dev/local/lib/python2.7/site-packages/mock/mock.py", line 1305, in patched
    return func(*args, **keywargs)
  File "/home/kirr/src/wendelin/z/zodburi/zodburi/tests/test_resolvers.py", line 258, in test_call_ipv6_no_port
    factory, dbkw = resolver('zeo://[::1]?debug=true')
  File "/home/kirr/src/wendelin/z/zodburi/zodburi/resolvers.py", line 141, in __call__
    host, port = netloc.split(':')
ValueError: too many values to unpack
---- 8< ----

Fix it by leveraging the fact that urlparse.urlsplit return() tuple-like object
with additional .hostname and .port attributes which already parse IPv6 correctly:

In [2]: u = urlsplit('http://[::1]:5553')

In [3]: u
Out[3]: SplitResult(scheme='http', netloc='[::1]:5553', path='', query='', fragment='')

In [4]: u.netloc
Out[4]: '[::1]:5553'

In [5]: u.hostname
Out[5]: '::1'

In [6]: u.port
Out[6]: 5553

In [7]: u = urlsplit('http://[::1]')

In [8]: u.hostname
Out[8]: '::1'

In [9]: u.port

In [10]: repr(u.port)
Out[10]: 'None'