Try to actually acquire a frame on Intel/Linux upon adapter inspection

[kvark]

Another take at the problem previously addressed by #94/#95/#97

[ConradIrwin]

@kvark I'm planning to pull this into Zed so we can get feedback from people.

[kvark]

Looks like this path faced some issues for end users

[apricotbucket28]

Hey, could this be rebased on master to fix #168? Thanks!

[kvark]

@apricotbucket28 done

[gretchenfrage]

Some information that may be relevant to this, especially if kvark finds it inconvenient to test Wayland locally:

• I still experience the issue described in https://github.com/zed-industries/zed/issues/17005 even if I switch Zed to use this branch.
• On that issue, I just added a backtrace of a segfault I captured with gdb, of encountering this issue. This may be useful as I didn't see a segfault already posted on related issue threads.

[kvark]

This path is no longer viable, since as of #203 we don't have a specific surface any more when creating a GPU context. We'll need to find a better path.