vcs: fix subprocessio process termination

659ecd26002c had a cut'n'paste error so it dropped process termination in the
SubprocessIOChunker close method.

vcs: fix subprocessio handling of hack for handling end-of-stream

Address pytype warning:
File "kallithea/lib/vcs/subprocessio.py", line 383, in __next__: No attribute 'stop' on listiterator[bytes] [attribute-error]

py3: officially support Python 3

All tests pass and no known regressions.

py3: use global next() function instead of .next() method

From 2to3 -f next.

py3: drop support for long - just use int instead

From 2to3 long.

py3: add b'' annotations in some places where they will be needed later

Mostly entirely trivial adding of b prefix that is a ignored for py2 ... and
also a bit of related trivial reformatting/refactorings.

vcs: drop subprocessio __del__ - it should no longer be necessary, and it might confuse both users and garbage collector

After 8dbe46ca608f, we always explicitly close so resources can be released
early.

__del__ makes it hard for the garbage collector to clean up, and it is
misleading to use it as if it was a reliable "Resource acquisition is
initialization" finale.

codingstyle: trivial whitespace fixes

Reported by flake8.

compat: drop unnecessary wrappers for old Python versions

Made unnecessary by other cleanup or upgrade to Python 2.6+.

subprocess: output on stderr doesn't mean that a process failed

Git sometimes does that. Threading non-determinism did that we sometimes didn't
see it, and sometimes we failed.

vcs: fix processing of git commands that return output on stderr

The subprocessio module used for interfacing with Git is using threading. It
might not be fully deterministic.

The subprocessio module also had the "feature" that it stopped reading and
reported failure once it found output on stderr - even if the command completed
(or would complete) with success.

Sometimes (like https://drone.io/bitbucket.org/domruf/kallithea/24 ) tests on
linux could fail on totally normal informational output like:

Couldn't run git command (['git', '-c', 'core.quotepath=false', 'checkout', 'foobranch']).
Original error was: Subprocess exited due to an error:
Switched to branch 'foobranch'

On Windows it would fail even more often.

To fix that, ignore stderr while processing output. There is a risk that it in
some cases can make the process block on stderr and thus never finish ... but
there is no reports of that yet.

git: run external commands as list of strings so we really get correct quoting (Issue #135)

a6dfd14d4b20 from
https://bitbucket.org/conservancy/kallithea/pull-request/17/add-quotes-to-repo-urls-for-git-backend
fixed that issue but did not make it "safe". The vcs git backend still used
command strings but tried to quote them correctly ... but that approach is
almost impossible to get right.

Instead, pass a string list all the way to the subprocess module and let it do
the quoting. This also makes some of the code more simple.

vcs: better handling of closed subprocessio InputStreamChunker source stream

test_git_cmd_injection would fail occasionaly.

cleanup: avoid some 'except Exception' catching - catch specific exceptions or log it and show what happened

This has a risk of introducing regressions ... but we want to get rid of all
exception muting and make the whole system less fragile and easier to debug.

vcs: close SubprocessIOChunker inputstream used for git (Issue #32)

git clone was leaking file descriptors

Second step in two-part process to rename directories.
This is the actual directory rename.