Increase the buffer pool size for GpuMemoryBuffer

When the system is extremely slow, the buffer allocation of the output
buffer can fail and upper layer would get the raw un-rotated buffer.

This CL mitigates the issue by increasing the pool size.

Bug: 1061346
Test: Run Hangouts Meet and `stressapptest -M 32 -m 0 -i 0 -c 2 -s 30`
concurrently, the flipped frame count in 30s dropped from 85 to 0 on
Nocturne.

Change-Id: I16a5cd260adc2799d8109bf323eae413ef7ac612
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2102268
Auto-Submit: Shik Chen <shik@chromium.org>
Reviewed-by: Ricky Liang <jcliang@chromium.org>
Reviewed-by: Guido Urdaneta <guidou@chromium.org>
Commit-Queue: Guido Urdaneta <guidou@chromium.org>
Cr-Commit-Position: refs/heads/master@{#750504}

services/video_capture/device_media_to_mojo_adapter.cc

@@ -181,12 +181,12 @@ int DeviceMediaToMojoAdapter::max_buffer_pool_buffer_count() {
 
 #if defined(OS_CHROMEOS)
   // On Chrome OS with MIPI cameras running on HAL v3, there can be three
-  // concurrent streams of camera pipeline depth ~6. We allow at most 24 buffers
+  // concurrent streams of camera pipeline depth ~6. We allow at most 30 buffers
   // here to take into account the delay caused by the consumer (e.g. display or
   // video encoder).
   if (base::CommandLine::ForCurrentProcess()->HasSwitch(
           switches::kVideoCaptureUseGpuMemoryBuffer)) {
-    kMaxBufferCount = 24;
+    kMaxBufferCount = 30;
   }
 #endif