rtpav1: Doesn't work with SVT AV1 encoder

$ gst-launch-1.0 --no-position -v videotestsrc ! video/x-raw,width=640,height=360 ! svtav1enc speed=10 ! av1parse ! rtpav1pay ! rtpav1depay ! fakesink
[...]
Pipeline is PREROLLED ...
Setting pipeline to PLAYING ...
Redistribute latency...
New clock: GstSystemClock
ERROR: from element /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0: Could not read from resource.
Additional debug info:
net/rtp/src/av1/depay/imp.rs(280): gstrsrtp::av1::depay::imp (): /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0:
Failed to read OBU header: forbidden bit in OBU header is set
Execution ended after 0:00:00.000426235
Setting pipeline to NULL ...
ERROR: from element /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0: Could not read from resource.
Additional debug info:
net/rtp/src/av1/depay/imp.rs(280): gstrsrtp::av1::depay::imp (): /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0:
Failed to read OBU header: forbidden bit in OBU header is set
ERROR: from element /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0: Could not read from resource.
Additional debug info:
net/rtp/src/av1/depay/imp.rs(280): gstrsrtp::av1::depay::imp (): /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0:
Failed to read OBU header: forbidden bit in OBU header is set
ERROR: from element /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0: Could not read from resource.
Additional debug info:
net/rtp/src/av1/depay/imp.rs(280): gstrsrtp::av1::depay::imp (): /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0:
Failed to read OBU header: did not expect size field
ERROR: from element /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0: Could not read from resource.
Additional debug info:
net/rtp/src/av1/depay/imp.rs(280): gstrsrtp::av1::depay::imp (): /GstPipeline:pipeline0/GstRtpAv1Depay:rtpav1depay0:
Failed to read OBU header: forbidden bit in OBU header is set
Svt[error]: deinit called without sending EOS!
Freeing pipeline ...

This might be a bug in the encoder (creating invalid OBUs), in the payloader (creating invalid packets) or the depayloader (parsing the data wrongly). Needs some more analysis.

#244 (closed) might also "solve" this by getting rid of most of the parsing in the depayloader, but then we'll have to see if decoders are happy with the output.