Fix race with short-lived TextureImageTransportSurfaces

Fix NULL ptr with short-lived TextureImageTransportSurfaces

When the surface is created from GpuCommandBufferStub, it is
made current for the very first time while
helper_->stub()->decoder()->GetGLContext() returns NULL (see order
in GpuCommandBufferStub::OnInitialize()).

If we then handle drawing and swapping of a frame while
a) the context is still current and we avoid calling
Surface::OnMakeCurrent() again
b) the surface is destroyed right after with the ack pending,
it is possible for context_ to be NULL in BufferPresentedImpl.

With virtual contexts, this patch causes us to hold a reference
to the real GL context instead now, which is fine because we only
need the context to delete the backbuffer texture during destruction.

BUG=169429
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=182403

[no sievers, 2013-02-14 01:23:34]

ptal

[piman, 2013-02-14 02:09:07]

lgtm

[commit-bot: I haz the power, 2013-02-14 02:42:14]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/sievers@chromium.org/12252028/2001