Code review comment for lp:~mir-team/mir/vsync-timings-from-server

Now tested manually (Nexus4 nested and non-nested), but needs work...

(1) VsyncProvider: name could be better. VsyncSource? But see also (4)

(2) Some very important regression tests have been deleted. I suspect it's possible to retain (convert?) them, so we should:
tests/unit-tests/client/input/test_android_input_receiver.cpp:
2488 -TEST_F(AndroidInputReceiverSetup, slow_raw_input_doesnt_cause_frameskipping)
2536 -TEST_F(AndroidInputReceiverSetup, rendering_does_not_lag_behind_input)

(3) Nested input (fingerpaint -w) is visibly more laggy with this branch. That probably would have been shown by the test cases (2) had they not been deleted.

(4) Input event spacing is significantly less smooth and irregular (both nested and non-nested) with this branch. Same kind of irregularity pictured in bug 1373692. Also the same kind of irregularity I encountered when I experimented with a virtual vsync timer (when the client receives the buffer). So there's presently no visible advantage to getting timings from the server over what you can calculate within the client.