refactor cli and validate before create rft

[xzrderek]

Note

Add dataset/evaluator pre-validation and Docker controls to RFT creation, extract shared CLI utils, and update local-test/upload plus tests accordingly.

• CLI (create rft):
  • Add flags: --skip-validation, --ignore-docker, --docker-build-extra, --docker-run-extra.
  • Refactor into helpers to resolve evaluator/dataset, validate locally, upload dataset, ensure evaluator, and create job.
  • Enforce dataset source rules: reject data_loaders JSONL; prefer input_dataset; fallback to detected dataset builders.
  • Validate dataset rows against EvaluationRow schema (sampled) before upload; optionally run selected test via pytest/Docker for evaluator validation.
• CLI (local-test):
  • Extract shared run path via run_evaluator_test; support Docker extras; improved error messages for multiple Dockerfiles.
  • Use shared utils for test discovery and entrypoint building.
• CLI (upload):
  • Use shared utils for discovery, ID normalization, entrypoint, account resolution, and dashboard URL building.
• Shared utilities:
  • New cli_commands/utils.py with test discovery/selection, evaluator ID normalization, account resolution, dashboard URL/link builders, dataset ID builder, entrypoint builder, and helpers.
• Tests:
  • Add/expand tests for RFT validation paths (dataset schema, evaluator pytest/Docker), dataset inference precedence, Docker flag propagation, and local-test behavior; update fixtures and helpers.

^{Written by Cursor Bugbot for commit 4a25e45. This will update automatically on new commits. Configure here.}

[dphuang2]

big PR, i'm assuming its just reorganization