Add --clean to avoid loading ftplugins

[nfischer]

This uses vim's '--clean' switch to avoid loading ftplugins, since this
can affect behavior during vroom tests. This switch seems to avoid the
issue in my local testing, whereas other things ('-i NONE',
'--noplugin', etc.) don't seem to have the same effect.

I can't get vroom running in '--neovim' mode (nvim 0.4.4), so this doesn't
attempt to address the issue in the neovim code path. neovim appears to
define '--clean' differently from vim, so it's possible neovim may
require a different fix.

I verified this still respects vroom's '--vimrc' switch while having the
desired behavior of avoiding plugins.

Fixes #105

[nfischer]

@dbarnett This seems to work correctly for vim 8.1.2269 (the particular test case I had issues with passes, as do all the other vroom test cases I'm aware of). Can you advise on how to check availability of the '--clean' switch? Is there an easy way to check the vim version?

[dbarnett]

Can you advise on how to check availability of the '--clean' switch? Is there an easy way to check the vim version?

I would suggest to always attempt to start with --clean but on errors fall back to running without --clean. That way in the common case there's no overhead.

Alternatives would be to always run vim --version first (guaranteed overhead, and could have gotchas trying to check that way anyway), or to accept new --clean and --noclean args in vroom that the caller has to configure correctly (awkward for users and still takes effort to implement).

[nfischer]

PTAL. It looks like the default wait time is 500ms which is more than enough for vim to recognize unknown switches. I tested this locally by changing --clean to --someunknownarg (to force this error case) and the fallback code worked as expected 10/10.

[dbarnett]

FYI, I'm thinking of eventually plumbing through and posting a warning notice in the final output if we couldn't use --clean.

[dbarnett]

That'll do the trick! Thanks!

Merging.