Add bingyu filtered ANN task submission

[Cooki-e]

Summary

Submission paths (new tasks): task-submissions/bingyu/1-x-1

Category / Implementation or Optimization / CPU or GPU: 1 / Implementation / CPU

Authors and coauthors: bingyu

Adds a Faiss metadata-filtered ANN retrieval task based on YFCC-10M.
The task includes 10 public queries and 10 hidden queries. Hidden queries are
scored individually, and their scores are averaged.
Index building has a 600-second deadline. Once the service reports ready,
each query has a 15 ms deadline.
The verifier checks result correctness, index persistence, and consistency
after index relocation and reloading, and performs an independent trajectory audit.

Validation evidence

Completed checks:

• python -B scripts/check_submission.py task-submissions/bingyu/1-x-1
  Result: 0 errors; submission review-stage static checks passed.
• python -B scripts/check_release.py
  Result: 0 errors, 0 warnings.
• python -B scripts/download_assets.py --task-path task-submissions/bingyu/1-x-1 --verify-only
  Result: size and SHA-256 verification passed for all 9 data files.
• git show --format= --check HEAD
  Result: no whitespace errors in the commit.
• Grader regression tests: all 13 tests passed after removing the separate
  service loading timeout.
• Real subprocess failure tests: timeouts, malformed JSON, and unexpected
  process exits were handled as expected. Injecting 3 failures among 10 queries
  produced a retrieval score of 0.7, with subsequent queries still evaluated.

Earlier full-corpus reference implementation validation:

• All 10 hidden queries passed across both runs, including the run after index
  relocation and reloading: 20 requests in total against the full 10-million-vector corpus.
• Index building took approximately 120.74 seconds. The slowest query took
  approximately 5.67 ms. The retrieval score was 1.0.
• This run preceded the final change removing the separate service loading
  timeout. The 13 regression tests passed after that change, but the full-corpus
  reference implementation validation has not been rerun.
• The runtime summary is available in the
  task README.

Development dataset: Cooki-e/search-swe-development

Contributor checklist

• All new tasks use exactly one contributor first-name namespace; temporary ordinals are category-unique and never reused after promotion; no direct formal additions or guessed final IDs.
• Actual authors in task.toml; commit email associated with my GitHub account.
• Allow edits from maintainers; promotion stays in this same PR.
• No secrets, downloaded inputs/models, job outputs, or author-only solutions.
• Static package/submission checks and applicable runtime checks recorded.
• Asset provenance, redistribution rights, licenses and pinned development SHA documented.

Maintainer merge checklist (new tasks)

• Design, verifier isolation and runtime evidence reviewed; exceptions explicitly approved.
• Updated to latest main; every final number assigned serially without collision.
• Separate pure git mv and reviewed finalization commits per task, in this PR.
• Per-task HF community PR or maintainer mirror merged into official dataset; manifest,
  SOURCES and license updated, all old assets preserved; official SHA pinned in assets.json.
• Repository inventories/docs updated; full checks and merge-ready gate pass;
  every task is promoted and no submission task.toml remains.
• Merge commit only (no squash/rebase); contributor history retained.