Improve range planning and immutable write scheduling

[Jason-Doyle]

Summary

• apply every supported scalar bound on the selected range index before loading candidates
• overlap independent immutable Snapshot or Trie writes with bounded secondary-index uploads while preserving HEAD as the final awaited compare-and-swap
• add protocol-equivalence, concurrency, publication-order, and range-candidate regression coverage
• update query, architecture, benchmark, and changelog documentation

Evidence

• the range regression covers 10,000 indexed documents and narrows a 25-document bounded result from the prior 7,500 candidates to 25
• the isolated seven-region experiment on experiment/parallel-write-pipeline measured roughly 15-21% lower large two-index write latency without changing stored keys, decoded object bytes, operation counts, or HEAD publication semantics
• absolute experimental p50 remained 5.67 seconds for Snapshot and 6.49 seconds for Trie, so this change does not claim sub-second writes

Verification

• npm run check
• npm test
• npm run verify:docs
• npm run examples:check
• npm run benchmark:compare:local
• npm run site:build
• npm run site:deploy:check
• npm run test:e2e
• npm run site:test
• npm run build
• npm run verify:package
• npm run verify:scaffold