@201 Add RO-Crate entity filters for targeted lookups

[arjlai221]

Summary

This MR fixes the RO-Crate existence-check path by extending the existing
GET /workflows/{id}/ro_crate_entities API with optional filters for:

• file_id
• entity_id

Instead of repeatedly fetching the full RO-Crate entity list and scanning it
client-side, callers can now perform targeted lookups against the server.

Problem

The previous implementation relied on listing all RO-Crate entities for a
workflow and then filtering in client code to determine whether a matching
record already existed.

That has two problems:

• it does unnecessary work for large workflows
• it does not scale well for repeated existence checks

What Changed

Server / API

• Added optional file_id and entity_id query parameters to
  GET /workflows/{id}/ro_crate_entities
• Plumbed those filters through the live router, server contract, HTTP server,
  and RO-Crate transport
• Updated the RO-Crate list query to apply those filters in SQL
• Applied the same filters to the count query so pagination metadata stays
  consistent

Client

• Updated the generated Rust client to pass the new query parameters
• Updated the checked-in Python and Julia generated clients for parity
• Added Rust helper methods for filtered lookup:
  • list_ro_crate_entities_with_filters
  • find_ro_crate_entity_by_file_id
  • find_ro_crate_entity_by_entity_id
• Switched find_entity_for_file() to use a targeted filtered lookup instead of
  full-list fetch + scan
• Added warning behavior if the Rust find_* helpers observe multiple matches
  and return the first result

Data Integrity

• Added a migration creating a partial unique index on
  (workflow_id, file_id) where file_id IS NOT NULL
• Migration deduplicates existing conflicting (workflow_id, file_id) rows
  before creating the unique index

Error Handling

• Duplicate RO-Crate inserts for the same workflow/file link now return a
  structured 422 response instead of falling through as a generic server error

Tests

Added and strengthened RO-Crate coverage for:

• filtering by file_id
• filtering by entity_id
• combined filters
• empty / non-matching filtered results
• pagination metadata on filtered listing
• duplicate workflow/file link rejection with explicit 422 assertion

Validation

Ran successfully:

• cargo fmt --all --check
• cargo check --tests
• cargo clippy --all --all-targets --all-features -- -D warnings
• dprint check

Note: full integration test execution is still limited in this environment
because sandboxed test server port binding is restricted.

Notes / Risk

• The new migration is intentionally data-shaping: if duplicate
  (workflow_id, file_id) rows already exist, it keeps the newest row before
  adding the unique index.
• The transport trait signature changed to carry the new optional filters, so
  any out-of-tree implementors will need to update accordingly.

[daniel-thom]

Delete?

[daniel-thom]

Why 2?

[daniel-thom]

Why do we need this?