[P1] Reduce Apple login provider round trips

[jjoonleo]

Summary

Fixes #323 by removing repeated Apple provider round trips from the returning Apple User login path and recording before/after benchmark evidence.

Root cause

Returning Apple Users still paid for two blocking Apple provider calls on every login request:

• Apple JWKS fetch during identity credential verification
• Apple auth/token exchange before the backend checked whether the Apple User already existed locally

That made the returning-user path depend on provider latency even when the local identity and existing User were enough to issue OnTime login tokens.

Changes

• Cache Apple JWKS for 24 hours and refresh once when a credential references an unknown kid.
• Check for the existing Apple User before exchanging the Apple authorization code.
• Skip Apple auth/token exchange for returning Apple Users.
• Keep Apple credential exchange for new Apple User registration.
• Add Apple login stage timing logs without logging sensitive values or sensitive key names.
• Update the Apple login benchmark harness so c10/c20 use one seeded returning Apple User per VU instead of racing the same User's active session.
• Fill docs/performance/apple-login-323.md with quick and bounded full before/after results.

Performance evidence

Full bounded run (RUNS=3 WARMUP_DURATION=30s MEASUREMENT_DURATION=60s):

scenario	before p95	after p95	JWKS calls/request	token exchange calls/request
c1 returning warm-cache	435.8 ms	46.5 ms	1.0 -> 0.0	1.0 -> 0.0
c10 returning warm-cache	429.1 ms	58.5 ms	1.0 -> 0.0	1.0 -> 0.0
c20 returning warm-cache	506.9 ms	60.4 ms	1.0 -> 0.0	1.0 -> 0.0

Quick and full summary CSVs are included under scripts/benchmarks/apple-login/results/*/summary.csv.

Validation

• ./gradlew test --tests devkor.ontime_back.global.oauth.apple.AppleLoginServiceTest --tests devkor.ontime_back.global.oauth.OAuthLoginFilterValidationTest
• ./gradlew check
• git diff --check
• git diff --cached --check
• scripts/benchmarks/apple-login/run.sh before quick
• scripts/benchmarks/apple-login/run.sh after quick
• RUNS=3 WARMUP_DURATION=30s MEASUREMENT_DURATION=60s scripts/benchmarks/apple-login/run.sh before full
• RUNS=3 WARMUP_DURATION=30s MEASUREMENT_DURATION=60s scripts/benchmarks/apple-login/run.sh after full