Read hot detection and cache management

[dongxinEric]

This is a sub task of issue #1250 and will be used to track all sub issues encountered.

[dongxinEric]

Todos:

Feature:

• [x] Log the read hot ranges when detected
• [ ] Cache management
• [ ] Right now the detection is happening on each individual storage server, and because each shard will be replicated across multiple storage servers, it could be possible to miss the case where each replica is approaching the read hot definition and FDB never reports a read hot, which is possibly not the correct behavior.
• [ ] Expose an API for client to query to get a list of read hot ranges happening currently, since the trace log may not be available to them.

Perf:

• [x] Minimize the performance cost of read detection #2376

Tests:

• [x] Add tests for the read sampling code to verify its correctness.