Adopt descriptive task names and update task authoring workflows

[hanhainebula]

Summary

Use stable, descriptive task names across task packages, evaluation commands, and Hugging Face data paths. Update contribution tooling and agent skills to guide authors through building, testing, and refining reproducible search-system tasks.

Changes

• Rename all nine task packages and synchronize their titles, configuration, asset manifests, and references. Pin migrated dataset assets to revision 6713b1eb7474e0768e906a0151c76af92f0cfcdd.
• Update launchers, downloaders, validators, and regression tests to use task names. Describe tasks through their objectives, interfaces, metrics, and resource requirements.
• Organize submissions as task-submissions/<task-name>, preserving the name through promotion to tasks/<task-name>. Update scaffolding, promotion, asset publication, and CI checks.
• Rewrite agent instructions as direct engineering requests while preserving structured specifications, submission contracts, scoring rules, and resource limits.
• Extend task-authoring guidance with configured coding-model trials, trajectory and score analysis, feedback-driven refinement, and validation of the final task revision.
• Require authoring agents to explain which API keys are missing and how contributors should configure them locally.
• Synchronize English and Chinese overviews, contribution guides, task documentation, and skill references with the current workflow.

Migration notes

• Commands now accept descriptive names, for example --task reasoning-query-rewriting. Numbered task identifiers are no longer accepted.
• New submissions use --task-path task-submissions/<task-name> for downloads and evaluation.
• Rename the local environment variable TASK_1_1_OPENROUTER_API_KEY to REASONING_QUERY_REWRITING_OPENROUTER_API_KEY, keeping its value unchanged.

Validation evidence

• Main regression suite: 177 passed, 2 skipped because the host lacks the required RewardKit version.
• Skill tests: 5 authoring and 10 maintenance tests passed.
• Task-local tests: 23 passed.
• Validated 9 Harbor configurations, 18 Compose overlays, 27 launcher previews, and 9 download previews.
• Verified all 335 migrated dataset asset references.
• Static package checks, merge-readiness and submission-history checks against the main base, skill validation, and whitespace checks passed.

Full container builds and new paid-model/GPU evaluations were not run for this migration.