perf(s3stream): fast path for block-start reads in StreamCache

[JasirVoriya]

Overview

• Adds an O(1) start-index resolution when startOffset equals the block’s startOffset in LogCache.StreamCache#searchStartIndex.
• Keeps the cache-hit path first; falls back to binary search only when neither cache nor block-start match applies.
• No behavior change for sequential, cursor-based reads.

Motivation

• First reads from a cached block often start at the block’s beginning. Previously, these took the binary search path.

Performance Impact

• Cold-start/scan reads from block start:
  • Avoids O(log N) binary search and a small helper allocation.
  • Improves latency deterministically for large blocks.
• Cursor-based reads:
  • Unaffected; cache hits remain O(1).

Committer Checklist (excluded from commit message)

• Verify design and implementation
• Verify test coverage and CI build status
• Verify documentation (including upgrade notes)

[superhx]

@JasirVoriya Thank you for your PR. From the perspective of LogCache's usage scenarios, most calls to searchStartIndex will not hit startOffset == this.startOffset, so I think this optimization may not be very helpful.