distutils scripts -- correct order of operations so 'from ... import' lines are correctly recognized.

My doctest / renormalizer fu is not strong enough to deal with the
changes to output:

- setuptools now has dependencies, and emits output to add them as
  'find-links'.  I tried adding an 'adding_find_link' renormalizer,
  but it doesn't suppress the output.

- Other failures due to merge tweaking APIs, etc.  Again, I'm not
  competent to fix them in the doctest / renormalizer world:  they make
  my head hurt.

This includes renaming `store_picked_versions` to `store_required_by`
to better signify its intent and adding `get_picked_versions`
to `zc.buidlout.easy_install`.

Fixes a bug with multiple import lines. We want to insert ourselves after the
*first* found import line, not the last one...

Also simplified an if/else for when there's no import statement at all.

Fixes #102

Distutils scripts can contain quite a lot of code. Up till now,
only a encoding hint was detected and placed at the top above
our 'import sys' and 'sys.path=' stuff.

Now we look for the first non-__future__ import line and place our
sys.path statement above that line. This makes sure the encoding hint,
the module docstring and the future imports are all above our code,
as should be.

There might be corner cases, but at least this fixes a couple of
common problems.

Fixes #82

Thanks Marius Gedminas for the fix I could copy/paste :-)

Note that there is no test for this as the test setup doesn't allow for it.
The error that this commit fixes, fixes an error that occurs at
easy_install.py import time, so the fix also has to be on that level, making it
hard to test.
I personally think it's ok like this... ;-)

When generating a python interpreter script with an empty `eggs=`,
previously you'd get `sys.path[0:0] = [,]`, which means a syntax
error because of the dangling comma.

os.umask is a bit weird in that it *does* return the current umask,
but it has a required argument that it immediately sets as the new umask...
Weird. Distribute solves that by assigning a dummy value and re-setting
the umask to the original value.

The matching test verifies that it actually works.

Note that I use python 2.6+'s octal notation: 0o755.
Buildout 2 doesn't support python 2.5 anymore, iirc, so that's fine.

A fixed (0755) execute permission is used in two places, I refactored it into
one method. Bonus: this makes it easy to test the fix proposed in
https://github.com/buildout/buildout/pull/18

Picked versions are stored for later printing/writing in the Installer
also when show-picked-versions is set to false but update-versions-file is set
to a file. Previously, show-picked-versions=false would also disable the versions
file updating.

Should never ever happen, of course, but somewhere in the universe exists a
distutils script with just the `#!bin/python` line... Added a len() to guard
against that.

Added a note as suggested by Marius

The second line is now inspected for a source encoding

I used Marius' suggestion (with one additional fix). The issue is demonstrated
by the test in @77cd8b16. This commit fixes
the failing test.

Fixes #64

Unlike the console-script entrypoint scripts, the distutils
scripts didn't get a -script.py appended to their name on
windows. This is now fixed.

I cannot handily use print_ from buildout.py as I get a circular import that way.

The Python package index is case-insensitive. Both
http://pypi.python.org/simple/Django/ and
http://pypi.python.org/simple/dJaNgO/ work. And distributions aren't always
naming themselves consistently case-wise. So all version names are normalized
and case differences won't impact the pinning.

Included: functionality (copy/pasted from buildout-versions) to
actually record and print the picked versions. For this, two new
class-level attributes on the Installer are added.

TODO: testing whether the printing of what required what also
works.

I need to go through buildout-versions' finish() method, still.

IncompatibleConstraintError.

       correctly.

  >=2.0dev in addition to being simple versions.

Maybe we can work with older versions of the egg recipe and other
recipes that biuild on it, making rollout of buildout 2 a bit simpler.

removing debugger, adding comments