fix: apply technology stack filter in recommendations (issue #1870)

[ionfwsrijan]

Problem

The Technology Stack dropdown on the recommendation page has no effect: the same recommendations are returned regardless of the selected stack. project_matches_tech is defined in src/utils/recommender.py but is never called, so the filter added in #936 and previously working via #1285 was silently dropped when the recommender engine was restored.

Fix

Apply project_matches_tech inside get_recommendations: when a non-all tech_stack is selected, filter the loaded projects before scoring, mirroring the original #936 implementation. Projects that do not match the selected stack are excluded from scoring, relevance, and final recommendations.

Files changed

• src/utils/recommender.py — filter all_projects with project_matches_tech when tech_stack is not "all".
• tests/test_recommender_tech_stack.py — new regression tests asserting the filter changes the result set, all returned projects match the selected stack, and filtered results are a subset of the unfiltered query.

Testing

• Isolated test run (seeded SQLite DB, app context): tests/test_recommender_tech_stack.py — 4 passed.
• Manual experiment against the real projects.json dataset confirmed the filter now narrows results correctly (e.g. java → 3 projects all matching, flask → 4 all matching, previously 30 unfiltered for every value).

Closes #1870

[vercel]

@ionfwsrijan is attempting to deploy a commit to the komalsony234-1530's projects Team on Vercel.

A member of the Team first needs to authorize it.