Protect PeerConnection lifecycle and JNI invocations with ReadWriteLock.

Add a ReentrantReadWriteLock in PeerConnection to prevent use-after-free
race conditions between JNI method invocations and dispose().
Concurrent calls to JNI methods while dispose() is executing could lead
to native memory corruption when freeOwnedPeerConnection is executed
while another thread is accessing the C++ PeerConnection instance.

Changes:
* Introduce lifecycleLock (ReentrantReadWriteLock) and wrap JNI-calling
  methods with callWithNative/runWithNative under readLock.
* Throw IllegalStateException("PeerConnection has been disposed.") when
  invoking methods on a disposed instance.
* In dispose(), execute close() and child disposals prior to acquiring
  writeLock to avoid deadlocks with signaling thread observer callbacks.
* Guard freeOwnedPeerConnection and zeroing nativePeerConnection under
  writeLock.
* Make dispose() idempotent so duplicate or concurrent calls complete
  safely.

Bug: b/533453798
Change-Id: If848b8dd7652c9810256414db890ec206cb8c996
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/500801
Commit-Queue: Linus Nilsson <lnilsson@webrtc.org>
Reviewed-by: Zoé Lepaul <zlep@webrtc.org>
Reviewed-by: Sergey Silkin <ssilkin@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#48530}