Fix concurrent sendMessage race by retrying on FAIL_NON_SERIALIZED

Replace tryEmitNext (fail-fast) with emitNext + busyLooping(100ms)
in StdioClientTransport.sendMessage().

The unicast sink's SinkManySerialized wrapper returns FAIL_NON_SERIALIZED
when two threads call tryEmitNext concurrently. busyLooping retries the
CAS spin instead of immediately failing, making concurrent sends safe.
The contention window is microseconds (single CAS operation), so the
100ms duration is just a generous upper bound for pathological cases
like GC pauses.

Before: 19/20 test repetitions fail
After:  20/20 pass

Closes #875

Author: AbbasNS

mcp-core/src/main/java/io/modelcontextprotocol/client/transport/StdioClientTransport.java

@@ -227,16 +227,17 @@ private void handleIncomingErrors() {
 
 	@Override
 	public Mono<Void> sendMessage(JSONRPCMessage message) {
-		if (this.outboundSink.tryEmitNext(message).isSuccess()) {
-			// TODO: essentially we could reschedule ourselves in some time and make
-			// another attempt with the already read data but pause reading until
-			// success
-			// In this approach we delegate the retry and the backpressure onto the
-			// caller. This might be enough for most cases.
+		try {
+			// busyLooping retries on FAIL_NON_SERIALIZED (concurrent tryEmitNext from
+			// another thread) instead of failing immediately. The contention window is
+			// microseconds (single CAS), so the spin resolves almost instantly; the
+			// duration is just a generous upper bound for pathological cases like GC
+			// pauses.
+			this.outboundSink.emitNext(message, Sinks.EmitFailureHandler.busyLooping(Duration.ofMillis(100)));
 			return Mono.empty();
 		}
-		else {
-			return Mono.error(new RuntimeException("Failed to enqueue message"));
+		catch (Sinks.EmissionException e) {
+			return Mono.error(new RuntimeException("Failed to enqueue message", e));
 		}
 	}