Optimize range cache performance and prepare v0.1.1

[xav-db]

What changed

• optimize bounded LRU metadata and sparse range traversal
• add fast paths for single-gap async reads
• strengthen deterministic concurrency, failure, property, and fuzz coverage
• add benchmark, coverage, fuzz, and Miri automation
• prepare range-cache v0.1.1 with Rust 1.86 as the MSRV

Impact

The public API and cache semantics remain unchanged. Sparse operations now scale with relevant ranges instead of all resident ranges, and common read-through paths perform less bookkeeping.

Validation

• cargo fmt --all -- --check
• core and async tests with default features disabled and all features enabled
• doctests with warnings denied
• Clippy across all targets and features with warnings denied
• Criterion benchmark smoke suite
• 196 cargo-semver-checks checks passed
• 100% source-normalized function, line, and region coverage
• cargo publish --dry-run for v0.1.1