Fix IoTConsensus batch accumulation and config reload - #18501

What changes were proposed in this pull request?

• Fix the batch-accumulation condition so the dispatcher waits after buffering the first request when write pressure is low.
• Make IoTConsensus configuration updates visible across dispatcher threads and propagate hot-reloaded limits to the existing dispatcher pipeline.
• Update the IoTConsensus queue memory ratio on configuration reload.
• Add unit coverage for batch accumulation, pipeline reload wake-up, and memory-ratio updates.

This is intentionally scoped to the lightweight, clearly incorrect paths; it does not claim to resolve every possible WAL/decompression performance issue.

Why are the changes needed?

The previous condition checked bufferedEntries.isEmpty() only after adding a request, so the accumulation wait was never reached. Existing dispatcher threads also retained stale reloadable limits.

How was this tested?

• mvn -pl iotdb-core/consensus clean test (72 tests, 0 failures, 0 errors)