[AubreyHaskett]

Summary

👋 RootEncoder exposes RTMP ping/pong RTT, but not the RTT tracked by the kernel for the underlying TCP connection, which I'd love to have. This PR will:

• expose the active java.net.Socket for direct RTMP connections that use the Java socket implementation
• surface the socket through RtmpClient, RtmpStreamClient, and GenericStreamClient
• return null when the connection is inactive, tunneled, uses another protocol, or uses a non-Java socket
• document that callers may inspect the socket but must not close or reconfigure it

Context

This is a follow-up to #2211. That proposal added native TCP_INFO support inside RootEncoder, which we do not want to do, as it required heavy non-Java/Kotlin additions to the repo.

This revision follows your recommendation -- RootEncoder only exposes its Java socket, while consumers can do what they like with that on their end (in my case, that will be obtaining a descriptor and implementing JNA/JNI diagnostics). This keeps code native, and keeps CMake/NDK config + ABI maintenance out of RootEncoder.

Users of RootEncoder will be able to use this socket access point for things other than kernel RTT as well (IP addresses/ports, correlating the connection w/app telemetry, additional info during SSLSocket use, additional TCP info, etc.).

Validation

• ./gradlew :common:testDebugUnitTest
• ./gradlew :rtmp:testDebugUnitTest
• ./gradlew :library:compileReleaseKotlin
• validated from the sample app with a test-only consumer using ParcelFileDescriptor.fromSocket() and getsockopt(TCP_INFO) during a live RTMPS stream: 111 consecutive RTT samples, 0 unavailable readings, 37.7–60.3 ms range, 48.0 ms average

Pedro - thanks for your time reviewing, chatting, and for the great library!

[AubreyHaskett]

Needed here and for the RtmpSocket so that every thread is told to read the latest value of the socket instead of potentially using a cached value. This is important for how I will be using the info from the socket, but should overall strengthen thread visibility without changing how the code behaves.

This will not do anything undesirable like change how the socket connects, sends data, reconnects, or closes.

[AubreyHaskett]

added because most (but not all) of the tests have this. hope that's okay.

[pedroSG94]

Thank you for the PR. Merged