feat: wisps bulk prediction launch

[vtnphan]

Pull Request

Summary

SBP-361 Add bulk-prediction workflow support to the WISPS pipeline, including dataset upload, workflow launch, and result settings display. Refactors interaction-screening–specific schemas and services into shared WISPS abstractions reusable across both workflows.

Changes

• New: bulk-prediction launch support — the launch_workflow handler now branches on "interaction-screening" | "bulk-prediction" and launches both via launch_wisps_workflow, using a WISPS_WORKFLOW_MODES lookup to set mode: g1-g2 or mode: manual in paramsText.

• New: unified dataset upload endpoint — POST /datasets/{workflow_name}/upload replaces the interaction-screening–specific endpoint. Accepts Literal["interaction-screening", "bulk-prediction"] as a path parameter and routes to the correct S3 base path (/g/data/yz52/sbp-service/input/interaction_screening or .../bulk_prediction).

• Refactored schemas — SequenceItem → WispsSequenceItem (group field is now optional to support bulk-prediction sequences without group); InteractionScreeningDatasetUploadRequest → WispsDatasetUploadRequest; InteractionScreeningFormData → WispsFormData.

• Refactored service — upload_interaction_screening_csv_to_s3 merged into single upload_wisps_samplesheet_to_s3(sequences, run_id, base_path, label). Branches on group field presence: interaction-screening CSVs include id, sequence, group, type columns (query→g1, target→g2); bulk-prediction CSVs include only id, sequence, type.

• Result settings tab — GET /{run_id}/settingParams now reads paramsText and configProfiles from queued_jobs.launch_payload (parsed via yaml.safe_load) instead of workflow_runs.submitted_form_data, so the exact values sent to Seqera are always shown.

• Dynamic WISPS mode — get_wisps_default_params now accepts an explicit mode argument; the executor derives mode from form_data.workflow via WISPS_WORKFLOW_MODES dict instead of hard-coding g1-g2.

• Tests and lint — updated all test imports (SequenceItem → WispsSequenceItem, etc.), added mode arg to get_wisps_default_params call sites in tests, applied ruff import-sort and black formatting across 17 files.

• No database migrations required for this change.

How to Test

1. Interaction-screening (regression) — submit an interaction-screening job as before; confirm the result settings tab shows paramsText with mode: g1-g2.

2. Bulk-prediction dataset upload:

   curl -X POST /api/workflows/datasets/bulk-prediction/upload \
     -H "Content-Type: application/json" \
     -d '{"sequences": [{"id": "seq1"}, {"id": "seq2"}], "runId": "my-run"}'

   Expect a response with s3Key, s3Uri, and splitOutputDir pointing to the bulk-prediction base path.

3. Bulk-prediction launch — submit a bulk-prediction workflow; confirm the Seqera launch payload contains mode: manual in paramsText.

4. Result settings — open any completed WISPS job's settings tab; paramsText should render as a structured object (not a raw YAML string), and configProfiles should be present.

5. Unit tests and lint:

   .venv/bin/python -m pytest          # 571 passed
   .venv/bin/ruff check .              # no errors
   .venv/bin/black --check .           # no reformats needed

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• I have added tests that prove my fix is effective or that my feature works
• I have added or updated documentation where necessary
• I have run linting and unit tests locally
• The code follows the project's style guidelines

[marius-mather]

looks good, just a couple of minor code cleanups to suggest

[marius-mather]

good to go, thanks!