IGNITE-28413 Java client: fix race condition in message decoder

[ptupitsyn]

Disable automatic discardReadBytes in Netty to fix a race condition and data corruption caused by concurrent buffer modification:

• Netty calls discardSomeReadBytes on IO thread, which changes readerIndex
• TcpClientChannel.asyncContinuationExecutor decodes the response on another thread, which also uses and changes readerIndex

https://issues.apache.org/jira/browse/IGNITE-28413

[Copilot]

Pull request overview

This PR addresses a Java thin client race condition in Netty message decoding where the same ByteBuf (and its indices/content) can be modified concurrently by Netty’s internal discard logic and by response decoding on another thread, leading to data corruption.

Changes:

• Make ClientMessageDecoder final.
• Disable (or attempt to effectively disable) Netty’s automatic discardReadBytes / discardSomeReadBytes behavior via setDiscardAfterReads(Integer.MAX_VALUE).
• Annotate the decode(...) override as @Nullable and remove the related suppression.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.