Multiple changes to make pylint happy (inc. fake mode changes)
    
    This is a large bunch of changes to resolve pylint issues (score
    goes from mid-5s to high-9s). For some of the issues pylint
    reported I went beyond just cleanups. Some apparently unused
    stuff was removed entirely, notably `TestSuite`'s results dict
    and `store_results` method, and its `in_output` method.
    
    The 'fake' environment handling was changed quite a bit to be
    substantially simpler. This involves changing the way it's
    invoked - there aren't two args any more, just a single arg for
    the release version; if it's set to a non-zero value fake mode
    is enabled, otherwise it isn't. This also removes the ability
    to run multiple scenarios including 'allinstall' and pass
    `--releasever` but not `--dryrun`, which would make the other
    scenarios run in 'real' mode but 'allinstall' run in fake mode.
    That just seemed like too much of a corner case to bother with;
    instead we just bail out if you try to include 'allinstall' but
    don't pass releasever.
    
    The fake root cleaning method is changed to a function as it
    doesn't need to be a method. Ditto for the stuff that creates
    the fake 'dnf module list' output, there's no need for that to
    be a class at all.
    
    Various methods that returned values were changed not to do so,
    when the thing that called them doesn't use the return values
    anyway.
    
    Lots of places were using format strings in logger calls, which
    you're not supposed to do. You're only supposed to use %-style
    formatting in logger calls. A pain, I know.
    
    There's also a lot of unused variables nuked, whitespace fixes,
    indentation fixes, comparisons like "if foo != None" improved,
    stuff like that.
    
    Signed-off-by: Adam Williamson <awilliam@redhat.com>