fix(wal): ensure WAL offset monotonicity under S3 backpressure

[Chandi977]

Bug Summary

Under S3 slowdown/backpressure, multiple concurrent calls to DefaultWriter.uploadBulk0() can reuse the same baseOffset, causing duplicate offsets. This violates WAL monotonicity guarantees and breaks Kafka-compatible behavior.

Steps to Reproduce

1. Pause MinIO (docker pause)
2. Run two high-throughput producers
3. Observe overlapping offsets in WAL logs

Expected

Offsets must increase strictly and never repeat.

Actual

Concurrent bulks reuse the same start offset under backpressure.

Maintainers: Please apply the labels bug, area/wal, severity/high.

I am preparing a PR with a fix + concurrency test.