test: core gameplay modules have zero unit coverage (puzzles, streak, quests, progress, categories, users) #664

Summary

Test coverage is skewed toward analytics/blockchain/support modules (~25 spec files), while the gameplay core is untested:

These services own XP math, streak state transitions, quest rollover, and puzzle retrieval - the logic players experience directly and analytics depends on. Regressions here are invisible until runtime.

Proposal

Start with the highest-risk pure-logic units, following the established mocking style (see score.service.spec.ts and challenge-attempt.service.spec.ts):

1. streaks.service / update-streak.provider: increment, same-day idempotency, date-boundary rollover.
2. create-puzzle.provider: DTO-to-entity mapping, difficulty enum handling.
3. users.service: uniqueness conflicts, findOneById contract used by JWT middleware validation.
4. Quest generation: deterministic day-key behavior.

Suggested split: one focused PR per module.

Acceptance criteria

• Each module above has its primary provider/service covered.
• New specs run in the CI test job.
• No behavioral changes to production code beyond bugfixes the tests surface.