Use video profile to decide if we should produce fp16 textures.

media/gpu/dxva_video_decode_accelerator_win.h

Index: media/gpu/dxva_video_decode_accelerator_win.h
diff --git a/media/gpu/dxva_video_decode_accelerator_win.h b/media/gpu/dxva_video_decode_accelerator_win.h
index 27f209bbee861e322ad7ab0d85baeb05516420f4..58b5cb7256237a8b7017b5958d1ddc159a8931e2 100644
--- a/media/gpu/dxva_video_decode_accelerator_win.h
+++ b/media/gpu/dxva_video_decode_accelerator_win.h
@@ -517,6 +517,9 @@ class MEDIA_GPU_EXPORT DXVAVideoDecodeAccelerator
   // Copy NV12 texture to another NV12 texture.
   bool copy_nv12_textures_;
 
+  // Copy video to FP16 scRGB textures.
+  bool use_fp16_ = false;
+
   // When converting YUV to RGB, make sure we tell the blitter about the input
   // color space so that it can convert it correctly.
   bool use_color_info_ = true;