Issue 9000: Provide parseable repr to datetime.timezone

Recently added datetime.timezone class does not have a custom repr, so one gets

>>> timezone.utc
<datetime.timezone object at 0x100681ef0>

instead of parseable

datetime.timezone(datetime.timedelta(0))

This is inconsistent with the way other classes in datetime module behave.

Merged issue5094 nosy list.

I am attaching a python prototype.  I am following the convention in place for the other datetime classes of prepending the repr with 'datetime.'.  This may not be ideal, but better than to introduce an inconsistency.  The only question I think is worth considering is the representation of timezone.utc.  I would like to have repr(timezone(timedelta(0))) == 'datetime.timezone.utc' even though currently timezone(timedelta(0)) and timezone.utc are different but equal objects.  This is going to change, however.  See issue9051.

[1] http://bugs.python.org/issue5094#msg106411

Attaching a C patch.

Replacing the patch.  The previously attached patch reverted issue9051 a bit too aggressively.

Committed in r82617.  Python version committed to sandbox in r82616.

messages: + msg109443
stage: commit review -> resolved

messages: + msg109438

stage: test needed -> commit review