See e.g. lha_os2_208/lfn.lzh or lha_os2_208/h3_lfn.lzh from
https://github.com/fragglet/lhasa/tree/master/test/archives.

This makes the code a tad clearer.

The output from the `lha` list archive command is eclectic and
inconsistent, and engrampa was actually crashing when opening certain
archives (eg. lha_os2_208/h3_subdir.lzh in the Lhasa test suite). But this
also fixes the list parsing more generally. With this change, engrampa
now appears to successfully open all of the .lzh files in the Lhasa test
suite correctly without crashing.

The first field in lha's list output either contains Unix permissions or
an OS name in [brackets]. There was already hard-coded support for
[MS-DOS], [generic], [unknown] and [Amiga], but other OS names were not
handled properly. This is now fixed, with an approach that is mindful of
the fact that the OS name can contain spaces.

Empty value (whitespace) in the UID/GID column is also handled
correctly.  This was partially working because of the aforementioned
special-casing, but now is fixed more generally.

This likely should not happen with valid dates, but can if a archiver
reports invalid ones for whatever reason (legitimate weird dates, or
parser failures for example).

RAR 7.0 outputs a different error when password is missing:

>
> Incorrect password for foo.rar
> Incorrect password for foo.rar
> Cannot find volume foo.r00

Whereas RAR 6.21 doesn't output the last 2 lines.

The error about volumes confuses the parser thinking there's a volume
issue, which hides the incorrect password one.

Note that RAR 7.0 outputs this *most of the time*, but once in a while
it'll also skip the last 2 lines.  Whether it is a race condition or
not, it at least happens often enough with RAR 7.0 to consider this its
normal behavior and handle it.

Fixes #526.

Fixes #527.

Fix crash if filename retrieval fails, which currently happens with
RAR 7.00 directories.

An upcoming fix with properly parse those, but it's better not to crash
on malformed/unexpected input nonetheless.

Part of #527.