Calculate the RMS audio level of audio packets being sent before
invoking an encoded frame transform, and pass them with the encode frame
object.
Before this, the audio level was calculated at send time by having rms_levels_ look at all audio samples encoded since the last send. This
is fine without a transform, as this is done synchronously after
encoding, but with an async transform which might take arbitrarily long,
we could end up marking older audio packets with newer audio levels, or
not at all.
This also makes things work correctly if external encoded frames are
injected from elsewhere to be sent, and exposes the AudioLevel on the
TransformableFrame interface.
Bug: chromium:337193823, webrtc:42226202
Change-Id: If55d2c1d30dc03408ca9fb0193d791db44428316
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/349263
Reviewed-by: Jakob Ivarsson <jakobi@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Tony Herre <herre@google.com>
Cr-Commit-Position: refs/heads/main@{#42193}
64437e8cc0