feat(blob): distribute blob sidecar keys with reversed binary ids

[Xuanwo]

This PR accepts suggestion from @jackye1995.

data_file_key already randomizes distribution across data files, but sidecars under a single data file still used sequential-looking names. Reversed-bit binary names improve prefix fan-out for this local sequence without introducing compatibility constraints (this version is unreleased).

• change blob v2 sidecar file naming from hex ({:08x}) to binary ({:032b})
• derive the file name from blob_id.reverse_bits() so early sequential IDs spread quickly across object-store key prefixes

---

Parts of this PR were drafted with assistance from Codex (with gpt-5.3-codex) and fully reviewed and edited by me. I take full responsibility for all changes.

[github-actions]

PR Review: feat(blob): distribute blob sidecar keys with reversed binary ids

No P0 or P1 issues found. This is a clean, well-scoped change.

The approach is sound: reverse_bits() on sequential blob IDs gives good prefix fan-out in binary representation, and the PR correctly notes that this is unreleased so there are no backwards-compatibility concerns.

One minor observation (not blocking):

• The starts_with("data") guard at cleanup.rs:464 is now subsumed by the stricter matches!(data_dir, Some(dir) if dir.as_ref() == "data") check. The matches! version is strictly better since starts_with("data") would also match hypothetical directories like "database/". You could consider removing the starts_with guard to avoid redundancy, but keeping both as defense-in-depth is also fine.

LGTM.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!