new patch-parser

[Harald Sitter]

I just deployed a new version of the patch-parser which is much more
complete and stuff:

http://qa.kubuntu.co.uk/~jr/patch-parser/

# changes
- no longer weirdly templates by modifying an html file (now uses haml
- an html template language)
- tracks more parts of git.debian: frameworks plasma kde-std
applications qt kde-extras (previously frameworks and plasma only)
- parses more branches: kubuntu_vivid_archive kubuntu_unstable
(previously only kubuntu_unstable)
- each branch creates their own html page with an index.html linking
to all generated branches
- dates are correctly picked up from upstream patches
- cached clones are stored in a subdirectory (./cache/)
- output is always sorted by repo name
- script currently will fail and not write anything if it encounters a
patch it does not know how to process (previously it would eat them
with nothing but a random message on stdout)

Unfortunately I also had to get rid of fallback git parsing. i.e. if a
patch is not dep3 its author and date will always be 'Unknown' where
previously the script would use data from git. This is in part because
touching git in this way is all sorts of bad responsibility isolation
and if we want to have this back it ought to be implemented in a
fashion abstract enough to also use support bzr.
That being said, bzr is still not supported, but much more feasible now.

There are plenty more things that could be improved such as unit
testing could be very helpful, alas since it is a chore to wire things
up to have sensible coverage on !github repos I hardly can be bothered
to. There's also the lack of bzr support which would also require
slightly more accurate cache ordering (currently we have a flat cache
./cache/$name which has all sorts of problems to it). The overall page
could also be improved and possibly changed such that it doesn't need
to point to anonscm but instead hosts all data local (e.g. the patches
are there in the clones anyway...).

I'll possibly venture into a change of the caching structure still
this week. Other than that I think the parser is in a pretty good
state now.

HS