docs: establish Phase 2 documentation baseline

[ZSeanYves]

Change summary

Govern all maintained documentation before Phase 2: remove the obsolete 0.7
migration guide, establish a documentation index and lifecycle, rewrite the
root/package/tool READMEs for the normalized src/ layout, update the
CHANGELOG and maintenance plan, and publish freshly generated formal benchmark
evidence.

Add a blocking standard-library documentation check for local links, required
and retired documents, byte-identical root READMEs, stale benchmark claims,
trusted/release-binary summary contracts, and evidence SHA-256 values.

Scope

• This PR is limited to one concern (behavior, refactor, dependency, performance, or governance).
• Non-goals are stated below.

Non-goals: product behavior, stable API, parser semantics, dependencies,
toolchain pinning, or approval of a new self-performance baseline.

Risk and ownership

• Risk: R2
• Affected formats/packages: documentation and governance only; benchmark summaries describe all formal rows
• Required owner(s): project and quality/performance owner
• Rollback version or procedure: revert commit 0a8e06c

Contract and compatibility

• Upstream MarkItDown behavior/vector, if applicable: formal oracle remains pinned to MarkItDown 0.1.7
• Surface compatibility impact: none
• Semantic compatibility impact: none
• Operational/security impact (network, external process, resource limits): no product change; documentation records existing optional-runtime boundaries
• API/CLI/error/golden changes: none; moon info produced no interface diff
• RFC/ADR (docs/rfcs/... or docs/adr/...) for stable API changes: not applicable

Performance

• Before/after command and platform: current-state full official-external-compare and official-self-baseline on Apple M4/16 GiB, macOS arm64, native release binaries, Python 3.11.15, MarkItDown 0.1.7
• Wall time / RSS / output size: external CLI median of row medians 63.941 ms and external reference 699.717 ms; self CLI median 52.947 ms and runner peak RSS 237,552 KiB; every evaluated CLI RSS budget passed with no missing measurement
• Why the output remains semantically equivalent: all 75 external tool cases and all 106 self CLI/engine cases were trusted; route, fidelity, provenance, density, semantic and RSS gates passed. The self run is explicitly a candidate observation because its fingerprints do not match the approved baseline.

Run IDs: run-1786101654079-0f0c773a82 and
run-1786102949457-9591fe380a.

Verification

Commands run:

moon info && moon fmt                         # 578 tasks; no interface work
moon fmt --check                              # pass
moon check --target all --warn-list +73 --deny-warn  # pass
moon test --target all                        # native 907/907; JS/Wasm/Wasm-GC 481/481 each
MOONBIT_NEW_NATIVE=1 moon test --target native # 907/907
moon build --target all                       # pass
python3 tools/governance/check_documentation.py # pass
python3 -m unittest discover -s tools/governance/tests -p 'test_*.py' # 10/10
python3 -m unittest discover -s tools/env/lib/tests -p 'test_*.py' # 29/29
find tools -type f -name '*.sh' -print0 | xargs -0 bash -n # pass
./tools/regression/check_coverage.sh --enforce # core 90.16%, formats 82.08%, tools 72.34%

• moon fmt --check
• moon info and expected .mbti diff reviewed
• moon check --target all --warn-list +73 --deny-warn
• Relevant native Tier 1 tests (macOS arm64/Linux x86_64)
• Contract/regression/coverage/security/performance lane as applicable
• Python/shell tooling tests as applicable

Fixtures, dependencies and generated files

• Fixture source/license/hash and behavior protected: no fixture changes; quality-lab remains pinned to d079e50b3e5ac181283c6c818e931bd2ef936a13
• Dependency/version/license/SBOM change: none
• Generated artifacts and regeneration command: two summary.json outputs copied from the named full release-binary runs under .tmp/bench/runs/<run_id>/results/summary.json; their SHA-256 values are recorded and enforced by bench/results/2026-08-07-macos-arm64/README.md
• Golden output explanation (required for every golden change): no golden output changes

Checklist

• I have not mixed an unrelated refactor into this PR.
• I added a regression test before changing an existing golden where possible.
• I updated capability/limitation, migration or changelog documentation when needed.
• I did not add Python/network/external-runtime requirements to core installation.