fix: prevent context timeout when messages delivered successfully

[anakys1997-stack]

Description

Implement double-check pattern in Writer.WriteMessages to prevent returning context.DeadlineExceeded when messages have been successfully delivered to the broker.

Problem

The issue occurred when both ctx.Done() and batch.done channels were ready simultaneously. Go's pseudo-random select could choose ctx.Done() even when the delivery promise held a successful response, causing:

1. WriteMessages returns context.DeadlineExceeded for messages that were actually delivered
2. Client applications execute retry logic, leading to duplicate records

Solution

The fix adds a non-blocking check on the batch.done channel before returning ctx.Err():

go select { case <-done: // Double-check if delivery completed before or concurrently select { case <-batch.done: if batch.err != nil { hasErrors = true } default: return ctx.Err() } case <-batch.done: if batch.err != nil { hasErrors = true } }

Testing

Added two unit tests:

• TestWriterWriteMessagesContextTimeoutWithSuccessfulDelivery: Verifies successful delivery is not lost
• TestWriterWriteMessagesContextCanceledBeforeDelivery: Verifies proper error when batch not delivered

Fixes #2