testcases for append entries hot spin

[codecasualty]

Fix:- WAL Bloat, Replication & Election Bugs

Problem

After ~43 hours of chaos testing, node3's WAL grew to 478 MB because:

1. Followers rewrote resent AppendEntries batches to the WAL even when there was no conflict.
2. The leader's replication loop hot-spun, resending unchanged batches ~4,400 times/sec.

This eventually caused WAL replay to exhaust the 512 MB heap and put the node into a crash-restart loop.

What Changed

Follower

• Implemented Raft §5.3 per-entry conflict detection.
• Skip entries already present with the same term or entries below the snapshot
• Only truncate and append the conflicting/new tail
• Added WAL durability barrier for retries
• Corrected matchIndex and commitIndex bounds

Leader

• Added a 20 ms backoff when nextIndex doesn't advance.
• Prevented hot-spinning on repeated successful responses.

Election

• A node becomes leader only if it is still a candidate in the same term.
• Prevents two leaders from being elected in the same term.

WAL Replay

• Replay is now index-aware.
• Duplicate/old WAL records are safely collapsed into a clean log.