fix: Add project filter to apply_data_source and delete_data_source

[mailtoboggavarapu-coder]

Summary

Fixes #6206

apply_data_source and delete_data_source in registry.py were not filtering by project, meaning an operation on one project could silently overwrite or delete data sources with the same name in a different project.

What's changed

• Added project-based filtering to apply_data_source: duplicate detection is now scoped to the current project
• Added project-based filtering to delete_data_source: the delete only targets data sources in the specified project

Tests

Added two regression tests in test_universal_registry.py:

• test_apply_data_source_cross_project_isolation: verifies applying to project A does not overwrite same-name source in project B
• test_delete_data_source_project_scoped: verifies deleting from project A does not affect same-name source in project B

---

[mailtoboggavarapu-coder]

Thanks for the detailed review, @devin-ai-integration! Fixed both encoding corruptions in commit f3437b08:

1. Line 706 — → (U+2192) was double-encoded as âÂÂ (bytes c3 a2 c2 86 c2 92); restored to proper → (UTF-8: e2 86 92)
2. Line 747 — same double-encoding issue; also restored to →

Both arrow characters are now correct UTF-8. The em-dash on line 700 was already fixed in the previous commit (3bcaf45).

[mailtoboggavarapu-coder]

Superseded by PR #6322 which has all commits properly signed with Signed-off-by: (DCO ✅). The DCO failure here was caused by the encoding-fix commits and the "Update branch" merge commit lacking sign-offs. Closing in favor of #6322.