Improve user-service availability-check latency under replay load

[kenahrens]

Summary

This PR addresses the user-service latency hotspot around availability-check endpoints (/api/users/check-username, /api/users/check-email) observed during replay analysis.

Changes

• add a short-lived in-memory cache (5s TTL) for usernameExists / emailExists checks to reduce repeated DB hits under bursty validation traffic
• invalidate relevant cache entries after successful user registration
• reduce log overhead on high-frequency availability endpoints by changing controller logs from info to debug
• make service/security log levels environment-configurable with safer defaults:
  • com.banking.userservice: ${APP_LOG_LEVEL:INFO}
  • org.springframework.security: ${SECURITY_LOG_LEVEL:WARN}

Findings (Issue #69)

Replay source

• backend/user-service/proxymock/recorded-complete

Environment comparison (before fix)

• local replay showed /api/users/check-username around ~42ms
• staging replay (do-nyc1-staging-decoy) showed /api/users/check-username timeout at ~10.01s with 100% failure for that endpoint in the sampled run

Local load comparison (before vs after this PR)

Replay command used for both runs:

• proxymock replay --test-against localhost:<port> --in proxymock/recorded-complete --times 30 --vus 5

Metric	Before (master)	After (this PR)
/api/users/check-username avg	4.33 ms	3.69 ms
/api/users/check-username p95	6.99 ms	6.00 ms
total requests	750	750
replay failures	0%	0%

Validation

• make test (user-service)
• local replay load benchmark before/after as listed above

Issue

• Closes Research: user-service high cpu and high latency #69