Reduce max time between repeat frame to avoid flakiness.

The frame cadance adapter injects repeat frames with a ca 1s interval
even if no new content has arrived, and the libaom AV1 encoder skips
higher temporal layer repeat frames for up to 1s. The causes a
potential race between the two that may in some edge cases result in
an extra keyframe being generated. Reducing the encoder interval to
500ms avoids that condition.

Bug: b/519058486
Change-Id: I3f84be1dbc5a2c160ce76afaa6034332f5429e10
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/485120
Auto-Submit: Erik Språng <sprang@webrtc.org>
Reviewed-by: Sergey Silkin <ssilkin@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#48078}

modules/video_coding/codecs/av1/libaom_av1_encoder.cc

diff --git a/modules/video_coding/codecs/av1/libaom_av1_encoder.cc b/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
index 380ff55..bedcbf1 100644
--- a/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
+++ b/modules/video_coding/codecs/av1/libaom_av1_encoder.cc
@@ -802,8 +802,9 @@
           it != last_encoded_timestamp_by_sid_.end()) {
         // Get the time since the last encoded frame for this spatial layer.
         // Don't drop enhancement layer repeat frame if last encode was more
-        // than one second ago.
-        if ((frame.rtp_timestamp() - it->second) > kVideoPayloadTypeFrequency) {
+        // than 500ms ago.
+        if ((frame.rtp_timestamp() - it->second) >
+            kVideoPayloadTypeFrequency / 2) {
           all_layers_droppable = false;
           break;
         }