feat(ocr): add built-in first-use capability

[ZhiXiao-Lin]

Summary

• add a3s-use-ocr as a reserved built-in CLI, standard MCP server, and content-bound Skill
• support bounded local Tesseract extraction and explicitly configured vision OCR with provider/data-egress diagnostics
• project canonical use/ocr, native Office, and optional Office compatibility routes for resident Code hosts
• add conservative MCP annotations across Browser, Office, and OCR so parent policy can distinguish local reads, mutations, and open-world calls
• package and smoke-test the OCR Skill in every release archive

This intentionally excludes the separately recovered Science toolkit and remote-extension registry work. It complements A3S-Lab/Code#14 without introducing a compile-time dependency on Code.

Safety boundaries

• no provider is installed or repaired implicitly
• local inputs are regular image files capped at 32 MiB and bound to a SHA-256 receipt
• non-loopback vision endpoints require HTTPS and an API key, and advertise off-device source transfer
• legacy OCR extension receipts cannot shadow the built-in route
• Office compatibility remains a distinct optional route; native Office stays provider-independent

Validation

• cargo fmt --all -- --check
• cargo test --workspace --all-features --locked -- --test-threads=1
• cargo test --workspace --all-features --locked
• cargo clippy --workspace --all-targets --all-features --locked -- -D warnings
• cargo metadata --locked --offline --no-deps --format-version 1
• git diff --check