Add TTL management and atomic counters to cache service

[NitinKumar004]

Summary

• Adds 7 new methods to cache driver interface: Expire, GetTTL, Persist, Incr, IncrBy, Decr, DecrBy
• Implements across all 3 providers: AWS ElastiCache, Azure Cache for Redis, GCP Memorystore
• Wires through portable API layer with metrics, recording, rate limiting, and error injection
• Incr/Decr on non-existent keys start from 0 (Redis behavior)
• Incr/Decr preserve existing TTL on the key
• GetTTL returns -1 for keys with no TTL (Redis behavior)
• Emits provider-native metrics for counter operations

Closes #64

Test plan

Unit tests (per provider - 8-12 tests each)

• TestExpire — set TTL on existing key, verify via GetTTL, verify expiry
• TestExpireKeyNotFound — returns error for missing key
• TestPersist — remove TTL, verify key survives past original expiry
• TestPersistKeyNotFound — returns error for missing key
• TestIncr — increment from 0, increment again
• TestIncrBy — increment by arbitrary delta
• TestDecr — decrement existing value
• TestDecrBy — decrement by arbitrary delta
• TestIncrNonInteger — returns error for non-integer values
• TestIncrPreservesTTL — TTL preserved after increment
• TestIncrCacheNotFound — returns error for missing cache (AWS)

Integration tests (cloudemu_test.go)

• TestCacheExpireAndPersistAWS / Azure / GCP
• TestCacheIncrDecrAWS / Azure / GCP

Verification

• Linter: 0 issues (golangci-lint run --timeout=9m ./...)
• Full test suite: all passing (go test ./...)