chore: release 0.7.0

[sylphx-desk-services]

What

Cuts lockdocs 0.7.0. The first-use benchmark work (#58), the free upgrade preview (#57) and lockdocs licence buy merged but were never released: npm @sylphx/lockdocs and the latest GitHub release are still 0.6.0, and the docs still showed the old 70/105 default first-use score.

• bun scripts/set-version.ts 0.7.0 and cargo update -w (same steps as the 0.6.0 release, chore: release 0.6.0 #53): all manifests, server.json, Cargo.lock at 0.7.0; CHANGELOG "Unreleased" becomes "0.7.0".
• Docs: default first use is 88/105 in 0.7.0 (home page proof, benchmarks page, README), above Context7's 77/105. The tables still hold the 0.4.0 run; the text says so and the next full benchmark run replaces them.

Release path

Merging to main runs release.yml (mcp-kit shared release workflow): native binaries, npm trusted publishing, npx smoke, GitHub release, MCP Registry. The crates job stays off until an owner sets CRATES_IO_PUBLISH_ENABLED.

Tests run

bun scripts/check-version.ts, check-tagline.ts, check-capabilities.ts pass; cargo metadata --locked passes. No Rust code changed.

Risks

Version-only change. The 88/105 figure is the local first-use run recorded in #58 (97 of 105 questions installable on that machine, held-out 14/16), not a CI run.

Readback for review

After the release workflow finishes: npm view @sylphx/lockdocs version is 0.7.0, the latest GitHub release is v0.7.0, then run the bench workflow on main and confirm first-use-default is at least 88/105 (the floor in bench/run.py).