Exclude -gnullvm target from i686_gnu build.rs

[jeremyd2019]

This was missed in #2774 even though the similar check was added to x86_64_gnu, presumably because there was no handling for i686-pc-windows-gnullvm at that time.

Follow-up to #2961 (comment)

[riverar]

Before we merge this in, we'd like to fix the current cross.yml to catch this error condition first. Then we can use that failing test to validate this new PR.

[jeremyd2019]

I'm not seeing how you would test that, but OK. Theoretically, if you had a branch where you reverted #2961 (except for the change to cross.yml), and applied the change from this PR to it, the i686-pc-windows-gnullvm check would fail. Then #2961 would be the "correct" fix for that failure...

Sorry for doing things kind of out of sequence here... I didn't notice the "oversight" from #2774 until I had already created #2961 (the failure of adding i686-pc-windows-gnullvm to check.yml to fail as expected was a bit of a mystery to me until later, as you could probably see from the comments).

[kennykerr]

Really appreciate you helping out here and debugging this further. I guess what I'm trying to say is that these checks don't appear to be necessary any longer and #2973 seems to confirm.

[jeremyd2019]

Yes, I figured this was an intentional "belt-and-suspenders" type situation, excluding both via the Cargo.toml dependencies and via the build.rs. But it would certainly make life easier if the targets were only filtered in one place.

[kennykerr]

Closing in favor of #2973.