Fix lost context recreation race in BrowserGpuChannelHostFactory

Fix lost context recreation race in BrowserGpuChannelHostFactory

An existing GPU process may have died between the last message we sent on its
channel and the next time we try to establish a new GPU channel. In that case,
we would re-use the GpuProcessHost but fail on the Send.

When that happens, we want to retry with a fresh GPU process.

BUG=129067
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=149993

[piman, 2012-08-03 22:16:55]

Note: I suspect we want some similar logic in GpuMessageFilter, though the race
is much less likely and not quite as problematic (failure to create the GPU
process on Aura forces us to kill the browser, whereas the renderer can survive
without one).
I will take a look at that later (and hopefully share some of the logic).

[apatrick_chromium, 2012-08-03 23:59:16]

lgtm

[commit-bot: I haz the power, 2012-08-04 00:17:54]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/piman@chromium.org/10852012/1

[commit-bot: I haz the power, 2012-08-04 01:27:08]

Change committed as 149993