media/gpu/v4l2ip: allocate/destroy buffers in device thread

We want to switch V4L2IP to V4L2Queue, but V4L2Queue is more strict
about all device access being done from the device thread. Currently
buffer allocation and deallocation is being done in the client thread,
so move them to be performed by the device thread instead.

BUG=792790
TEST=Checked that VDA passed on Hana and VEA on Peach

Change-Id: I4b06b3bb10676c3a902e5e91657da2ecca7b737c
Reviewed-on: https://chromium-review.googlesource.com/c/1071496
Commit-Queue: Alexandre Courbot <acourbot@chromium.org>
Reviewed-by: Hirokazu Honda <hiroh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#629119}
2 files changed