ci: remove some freedreno a6xx skips

I pushed a CI run and restarted the a630 gles3/gles31 jobs 3 times (for a total of 4 runs each). I did see one dEQP-GLES31.functional.layout_binding.ssbo.fragment_binding_array flake so I added that one back to the skips. I'll try restarting them a few more times over the weekend since I guess the runners won't be heavily loaded (ie. if I happen to walk by my computer and notice the previous run finished I'll start it again).

I wonder what is a reasonable procedure/threshold to de-list things that were flakey at one point, and at least seem not to be now? Maybe just remove them from skips and watch #freedreno-ci to see if they start showing up again? @anholt?

I guess it would be nice if we separated flakes into their own list, and ran them in CI but allowed Failed status on those tests to not fail the CI job. With some tracking of flakes over time we could maybe even correlate increases/decreases to driver changes.