Minor Dtls-in-STUN-encapsulation cleanups

That unfortunately does not solve the existing bug :(

1) Add a new method GetPending that is used when retransmitting
unencapsulated, this only matters for PQC and means that we might
send 2 packets "back-to-back", which is how it's ordinarely done.

2) Add a new UpdateHandshakeTimeout, that is used after the initial
estimate and pipe that all the way to
rtc_base/openssl_stream_adapter.cc.

3) Move the RunBenchmark into the test fixture, that way it can easily
be used from other tests.

4) Modify the "with packetloss" test to run with 25% packetloss, from 50%. This might seem like a bummer, but at 50% the tests very super fragile, changing *anything* made them randomly fail. I.e. we had overfitted the settings *a lot*. At 25% they are stable.

5) Add a new test that verifies that using dtls-in-stun leads to short
connect time that not using it.

BUG=webrtc:367395350

Change-Id: I9cc207eab413dc0e7542464174e504a39f425466
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/435182
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#46524}