While invoke setup.py certain eggs (like scikit-learn) launch
cetain custom builds (for cython) using subprocess and sys.executable.

This commit aims to preserve the sys.path over the runs, even if an
egg is using subprocess with the same python to build a component of
the egg.

- Support on the fly patches in zc.recipe.egg by ``EGGNAME-patches``,
  ``EGGNAME-patch-options``, ``EGGNAME-patch-binary`` (or
  ``patch-binary``) and ``EGGNAME-patch-revision`` options.

- Support on the fly patches in zc.recipe.egg:custom by ``patches``,
  ``patch-options``, ``patch-binary`` and ``patch-revision`` options.
  (options ``EGGNAME-*`` are also supported as well).

Specified patches are automatically applied on required eggs as well.

This fixes cache of patches.

Clean-up + fix issue found at nexedi/slapos!1674

Currently only zc.recipe.egg:custom supports setting environment
variables, and zc.recipe.egg:develop does not.

My motivation for allowing setting environment in :develop is
wendelin.core

    https://lab.nexedi.cn/nexedi/slapos/blob/b5faab3b/component/wendelin.core/buildout.cfg

There we have [wendelin.core] part which installs released egg from
pypi, and [wendelin.core-dev] part which installs wendelin.core from
its latest git version via zc.recipe.egg:develop .

The problem is, wendelin.core for setup.py to work, needs git available,
and with slapos we usually don't have git available on base system, so
we build it by our own and do something like

    [wendelin.core-dev]
    recipe = zc.recipe.egg:develop
    environment = wendelin.core-dev-env

    [wendelin.core-dev-env]
    # wendelin.core-dev needs git to build
    PATH = ${git:location}/bin:%(PATH)s

and the problem is environment does not currently work for
zc.recipe.egg:develop, and thus git is not found -> build fails.

~~~~

In order to support environment in :develop, we just move environment
setting/restoring bits from Custom to Base, and provide Base.install() which
uses this bits. Custom & Develop .install() becomes ._install() which gets
hooked into Base.install() .

I've tested the patch only manually, because currently automated tests are
broken in a lot of places for slapos.buildout and zc.recipe.egg .

/cc @kazuhiko, @Tyagov

Before this, sys.exit(1) was called to terminate the program directly.

To be dropped once all buildout in the wild are able to upgrade to a
version that supports new names in expression of conditional sections
(see previous commit).

Adding new names for expression is currently not possible because
buildout aborts before it tries to upgrade (in-place or bootstrap).

For example, existing values were not enough to distinguish
'arm-linux-gnueabi' from 'arm-linux-gnueabihf'.

The 'multiarch' value is the output of

  $CC -dumpmachine

where CC defaults to 'gcc'.

See also https://wiki.debian.org/Multiarch/Tuples

(changed option or missing path)

Sometimes, most parts are reinstalled for a reason that the user
didn't think about and it can take time to understand why.
Explaining for all parts would be too verbose and useless because
many are reinstalled just because their dependencies changed.

Egg is installed from wheel when the requested version ends with :whl

Reuse more code from slapos.core to support case of parent directory
not writable.

When recipes mutate the options, we should not allow invalid syntax
otherwise it gets written in .installed.cfg and generate invalid
configparser syntax.

When recipe mutate options, options values are written to
.installed.cfg without escaping buildout substitution syntax,
so if a recipe sets an option value that could be interpreted
as a buildout substitution, it is written as is in .installed.cfg.

This can be a problem if options read from _read_installed_part_options
are accessed, like it's the case with slapos patched buildout which
saves installed options after an error with part installation or after
each part installation when running in verbose mode.

python ZipFile module does not support updating an entry in place, instead make
a new zip file and copy all entries.

This avoids rebootstrap changing the signature of zc.buildout when
the package is in sites-packages for the initial Python version but
not in the sites-packages of the target Python (when it's not the
same Python)

Dists in python library (e.g. site-packages) using .dist-info format
are marked as DEVELOP_DIST by setuptools. Use their version to sign
these, and not their location like actual develop dists.

To set a default requirement for not-yet processed parts

The version of Python should not affect the behaviour of a recipe.
Anyway, it was already ignored for DEVELOP_DIST eggs.

This makes the slapos mechanism to share parts more efficient. And with
the upcoming changes in buildout & slapos.recipe.cmmi, there would be
no way for the slapos.reboostrap extension to prevent everything from
being rebuilt when reboostrapping to a different version of Python,
if bootstrap parts are shared.

The monkey-patch by slapos.reboostrap now becomes useless.

_install_and_load is slow, using cache saves time when there are
many sections.

Introduce a workaround for the shebang line length limitation
in zc.buildout.easy_install.

Useful when recipes generate non-string values to be reused by other
recipes.

Add referred parts' hash strings in __buildout_signature__, that
invokes rebuild of a part when one of its (recursive) dependencies
are modified.

Also remove duplicates and sort entries in __buildout_signature__.

Also, updating a part does not put it anymore at the end of the list of
installed parts, that was making .installed.cfg too big.

https://bugs.launchpad.net/zc.buildout/+bug/144228

Even though such configuration is wrong...

Before this, += and -= worked well when applied to keys defined in the
current section or in a section of the same name inherited from a file
extended by the current file, but not to keys inherited from another
section using macro extension <=, because <= substitution happens later
and all the += and -= are already resolved.

e.g.:

```
[macro]
a = 1

[part]
<= macro
a += 2
```

was equivalent to

```
[part]
a = 1
a = 2
```

instead of

```
[part]
a = 1
a += 2
```

Now a partial and brittle support for this is enabled by postponing
+= and -= resolution until <= substitution happens when the current
section contains a <= extension. But this is not guaranteed: if the
current section is overloaded in another file extending the current
one, then += and -= resolution will happend then, before <=.

A consistent solution would be to unify the implementation of
extends= and <= and update sections by taking both into account,
instead of updating based only on extends and computing <= later,
leading to inconsistencies. This could be achieved e.g. by computing
section updates on demand during substitution.
Co-authored-by: Xavier Thompson <xavier.thompson@nexedi.com>

Before this, += and -= worked correctly when applied to a key defined
in a file extended by the current file, but not when the key was in
the same file.

e.g.:

```
[part]
a = this case is
```

```
[buildout]
extends = a.cfg

[part]
a += ok
b = this case is
b += not-ok
```
Co-authored-by: Xavier Thompson <xavier.thompson@nexedi.com>

This test asserts buildout's behavior with regards to download options,
and this was changed by the new algorithm for extends. Tests for the
new behavior have not been written yet.

The new algorithm avoids fetching the same extended file more than once
and correctly handles overriding values and += and -=:

The new algorithm starts as if there was a buildout file containing

```
[buildout]
extends =
  user/defaults.cfg # if it exists
  buildout.cfg # if it exists
  command_line_extends.cfg # if passed on the command line
```

The files are then fetched in depth-first-search postorder and fetching
child nodes in the order given by the extends directive, ignoring files
that have already been fetched.

The buildout dicts are then collected in order, and this linearisation
is then merged at the end, overriding the first configs collected with
the later ones. The first dict in the linearisation is not from a file,
but the dict of buildout's (hardcoded) defaults. This is equivalent to
acting as though every file that does not extend anything extends these
defaults.

The first time a file must be downloaded from a url, the linearisation
is merged with the configs already collected, and the resulting options
are then used to determine the download options for this download, and
every subsequent download.

This is a break with buildout's current logic for download options.

By analogy with classes in Python, we are computing a linearisation of
the class hierarchy to determine the method resolution order (MRO).
This algorithm is not the same as Python's MRO since Python 2.3 (C3).

It could be good to switch to a C3 linearisation like Python.

This avoids unecessary copies. This is a preparatory step to
reimplementing the extends algorithm. It may be that this breaks
the extends algorithm as it is currently implemented.