[lanyue-llk]

Shorten benchmark release names to bench-vMAJOR.MINOR.PATCH and make the README a starting point for using the benchmark.

• Use bench-v1.0.0 for the historical report's scorer source and bench-v2.0.1 for the current 550-task catalog. Update the catalog manifest and its validator together; remove the old long-prefix aliases.
• Keep the task population, reference hashes and scoring source unchanged. Browser Eval publishes its own software versions as vMAJOR.MINOR.PATCH; reports bind both versions and exact code/input identities.
• Remove the repository file inventory, repeated ownership explanations and detailed scoring changelog from the README. Keep task suites, working examples and links to the owning guides, aligned in English and Chinese.
• Keep citations short and align CITATION.cff with Apache-2.0.

Validation: make check passed 2,579 tests and 321 subtests. The public version command verifies and reports bench-v2.0.1. README links and bilingual citations were checked.

Companion report build identities: lexmount/browser-eval#66.

[lanyue-llk]

Reviewed d0e78e7, including the tag validator/manifest change and the shorter English and Chinese READMEs. No new actionable defect found in this PR. The catalog tag and validator agree on bench-v2.0.1; task populations, reference bindings and scoring-source identity are preserved. The previous ordered-tree fingerprint omission is fixed in the base revision, and its reference/threshold regression controls pass.

Validation: make check passed 2,579 tests and 321 subtests. I also confirmed that a changed ordered-tree reference rejects a patch successor and fails relaxed release verification. The companion Browser Eval build-identity issue is reported separately on lexmount/browser-eval#66.