docs(computer-use): add Chinese counterparts for the Computer Use group

[wy471x]

Summary

The Computer Use document group under docs/ is English-only, while the rest of the
repository already publishes Chinese counterparts for its cross-cutting documents
(docs/architecture/*, docs/blogs/*, docs/cli-npm-release.zh-CN.md, …). This PR brings
the group onto that existing convention instead of inventing a new one.

Added — seven Simplified Chinese counterparts:

• docs/computer-use-model-loop-foundation.zh-CN.md
• docs/computer-use-evidence-classes.zh-CN.md
• docs/computer-use-provider-evidence.zh-CN.md
• docs/computer-use-host-events-contract.zh-CN.md
• docs/computer-use-provenance.zh-CN.md
• docs/computer-use-cursor-provenance.zh-CN.md
• docs/computer-use-ui-coverage.zh-CN.md

Changed — every English source of those seven gains counterpart: ./<name>.zh-CN.md
and moves from translation_status: source-only to synced; docs/README.md links the
Chinese versions from its Computer use section.

Deliberately not translated — computer-use-foundation-contract.md declares
language: zh-CN / source_language: zh-CN and its body is already Chinese prose, so it
gets no counterpart and its translation_status stays source-only.

The Chinese text follows the group's terminology policy: code identifiers, file paths,
commands, type/field/API names, and product or protocol names stay English, and every
inline identifier, numeric constant, threshold, commit SHA, digest, and enumeration value
(real-runtime, fault-injection, hermetic-protocol, skylight_pid, …) is copied
verbatim. dispatch, provider, and observation stay English to match the Chinese
foundation contract. Cross-document references point at the Chinese counterparts;
computer-use-provider-evidence.zh-CN.md links
./computer-use-evidence-classes.zh-CN.md#lab-fixture-准备.

The provenance documents are translated without changing claim strength: "provides
evidence, not a license grant", "no source code was copied into this repository", "not a
legal determination about the former facts", the "historical only" status of the
binary-recovered cursor inputs, and the MIT / © 2026 Leo attribution rows keep their exact
meaning.

Three translated documents (computer-use-provenance, computer-use-cursor-provenance,
computer-use-ui-coverage) are not listed in docs/README.md today, so they get no index
link here; that index gap is left as it was rather than fixed in a translation PR.

Verification

Claim	Evidence
Structure parity, all 7 pairs	a comparator over source and counterpart reports identical heading count/levels, byte-identical fenced code blocks, identical table dimensions, and identical list shape; 0 files with structural problems
Identifier and value parity	every inline code span in the English source appears in the counterpart; SHA-256 digest, commit SHAs, thresholds, and counts compared token by token
Cross-reference anchor	#lab-fixture-准备 resolves to the ### Lab fixture 准备 heading in the target file
Frontmatter	all 7 counterparts carry the same doc_id as their source, language: zh-CN, source_language: en, a working counterpart path, translation_status: synced, and the source's last_verified; each English source links back
npm run lint, npm run format:check	clean (3591 / 2122 files)
npm run check:asf-headers	the 7 new files carry the ASF header; the only reported failures are the pre-existing untracked .idea/ and .commandcode/ files, which are not part of this PR

Not run: npm run build, npm run typecheck, and the workspace test suites — this change
touches Markdown only and changes no code path.

Review focus

Reviewing a translation means checking semantic parity rather than sentence-by-sentence
fidelity: scope, guarantee strength, lifecycle status, numbers, and failure behavior must
match, while the Chinese prose is expected to read naturally on its own. The bilingual
checklist in skills/maka-architecture-docs/references/quality-gate.md is the standard
applied here.

AI use

Select exactly one:

• No generative tool made a substantive contribution
• Generative tooling made a substantive contribution

Tool(s) and scope: Not applicable — CONTRIBUTING.md excludes translation from substantive
contribution. The commit identifies its automated author with a Co-authored-by: CommandCodeBot <noreply@commandcode.ai> trailer.

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

The first item has no surface here: the diff is Markdown only, and the repository's gates
for documents are the ASF header audit and the biome lint/format run recorded above.
Typecheck and the workspace suites are unaffected because no code changed.

Does this PR entail a change in behavior?

• Yes — described under Summary above
• No