Encode key frame when enabling upper layer.

Request key frame when upper spatial layer is enabled dynamically
and inter-layer prediction is disabled or limited to key pictures.

This is needed to force encoder to produce RTP compatible bitstream
where temporal prediction is limited to the same spatial layer.

Bug: webrtc:9217
Change-Id: I4fc1e3f067689ba7b5c6bd1f5af922a0637f03d7
Reviewed-on: https://webrtc-review.googlesource.com/73580
Commit-Queue: Sergey Silkin <ssilkin@webrtc.org>
Reviewed-by: Rasmus Brandt <brandtr@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#23102}

modules/video_coding/codecs/vp9/vp9_impl.h

@@ -120,6 +120,7 @@ class VP9EncoderImpl : public VP9Encoder {
   const VideoFrame* input_image_;
   GofInfoVP9 gof_;       // Contains each frame's temporal information for
                          // non-flexible mode.
+  bool force_key_frame_;
   size_t pics_since_key_;
   uint8_t num_temporal_layers_;
   uint8_t num_spatial_layers_;