FAC-114 feat: add CSV test submission generator

[y4nder]

Summary

Backend API + admin console UI for generating realistic test questionnaire submissions with OpenAI-powered multilingual comments (Cebuano, Tagalog, English, code-switched).

Motivation

Manually constructing CSV files with realistic submission data for questionnaire ingestion is too slow for rapid analytics testing. The team needs volume (up to ~50 submissions per course) with realistic, multilingual qualitative feedback to properly exercise analytics dashboards (sentiment analysis, topic modeling, etc.).

Scope

API Endpoints

• GET /admin/generate-submissions/status — lightweight pre-check (enrollment/submission counts)
• POST /admin/generate-submissions/preview — generate rows with OpenAI comments
• POST /admin/generate-submissions/commit — submit via QuestionnaireService.submitQuestionnaire() per row
• 4 new filter endpoints: faculty, courses, questionnaire-types, questionnaire-versions

Admin Console UI

• Two-track selection form (faculty+course, type+version) with cascading dropdowns
• Auto-fetched submission status warning before generation (saves OpenAI tokens)
• Color-coded preview table with tooltips
• Commit result dialog with success/partial/failure states

Key Design Decisions

• Direct submitQuestionnaire() calls (bypasses ingestion pipeline to avoid DI scope issues)
• OpenAI gpt-4o-mini for comments with fallback on failure/timeout
• Per-student answer tendency for realistic distribution
• @ArrayNotEmpty() + @ArrayMaxSize(200) on commit rows
• Dangerous-key rejection for answers (__proto__, constructor, prototype)

Tests

• 31 unit tests across 3 suites (CommentGeneratorService, AdminGenerateService, AdminFiltersService)
• All 757 project tests passing

Tech Spec

_bmad-output/implementation-artifacts/tech-spec-csv-test-submission-generator.md