[electronicboy]

Exposes the three SCTP stats added in opencollab-incubator/libdatachannel#4 on SctpStats:

• sendQueueBytes(): bytes the association has accepted for sending and not had acknowledged, sent or not. DataChannel.bufferedAmount() only counts what usrsctp would not take yet, so up to a whole send buffer (1 MiB by default) waited below it unseen.
• sendBufferSize(): the bound on that queue, so a reader can tell how full it is and confirm SctpSettings.sendBufferSize took effect.
• receiveQueueBytes(): received bytes awaiting reassembly or ordering, the byte counterpart of pendingChunks().

Depends on opencollab-incubator/libdatachannel#4

• 3e6f753 syncs jni/libdatachannel to 85e65386, the commit in Report the bytes in the SCTP association's own queues and the send bu… libdatachannel#4.
• 130c1eb points .gitmodules at electronicboy/libdatachannel so that commit can be fetched meanwhile. It gets dropped once that PR merges, and the submodule is re-synced to the merged commit.

Testing

NativeTransportProbe now:

• sets sendBufferSize(512 KiB) and asserts sendBufferSize() reports exactly 524288;
• sends 32 × 16 KiB from the host and asserts the send queue holds at least one unacknowledged message, then drains to 0.

Run locally against an Mbed TLS build of the natives (no OpenSSL headers on that machine): all 15 probe scenarios pass, with peaks of 354–376 KB queued. The OpenSSL-dependent native test executables, the other probes and the release builds were not run locally.