Pkgcheck

--------------------

* addition of __attrs__ to base.Result classes; use this if __slots__ doesn't
suffice for listing the attrs to pickle.

* Thanks to Michael Sterret for pointing it out; tweak cleanup scan so that it
notes 1.12 overshadows 1.11 (stable keywords overshadow earlier unstable
versions): for example-
1.11: ~x86 ~amd64
1.12: x86 ~amd64

--------------------

--------------------

* treat pkg.restrict as a depset.

--------------------

--------------------

* drop digest specific checks; portage now prunes digests on sync regardless
of whether or not the repo is m2 pure; thus, no way to detect if a missing
digest is actually a screwup in the repo, or if it's portage being 'special'.
May re-add the checks down the line, currently however removing them for
the common case.

* back down check for files directory if manifest2; manifest2 glep didn't
specify that files directory could be dropped, but portage has deviated there.
Since been backed down, but getting ahead so we don't need an intermediate
release when they try it again.

* added check for missing metadata.xml; refactored common error class selection
logic into base class.

--------------------

--------------------

* correct tracebacks when dealing with a few result objects from repo_metadata

--------------------

--------------------

* makes StaleUnstable abide by --arches; ticket 59 (thanks leio).
* stop complaining about empty keywords, since they're now allowed instead of
using -\*.

------------------

------------------

* heavy refactoring of reporter subsystem, and clean up of check results.
Better messages, better output for normal usage. to_xml() methods were
dropped (XmlReporter handles it on it's own), same for to_str() in favor
of short_desc and long_desc attributes.
* whitespace checks now output one result for each classification for an
ebuild, instead of emitting reports for each line.
* all remaining 'info' statements are pushed to stderr now.
* new PickleStream reporter; used to serialize check results, and flush the
stream out stdout. If you need to get at the data generated, this is the
sanest way to do it (alternatives require trying to deserialize what a
reporter does, thus losing data).
* added new tool replay-pcheck-stream; used to replay a pickle stream through
alternative reporters.

------------------