feat(prometheus): add metrics for AI cache hits, misses, bypasses, and embedding latency

[janiussyafiq]

Description

Adds Prometheus metrics for the ai-cache plugin (#13578, #13632, #13644), integrated in prometheus/exporter.lua following the existing llm_* metric pattern:

• apisix_ai_cache_hits_total (with layer="exact"|"semantic"), apisix_ai_cache_misses_total, apisix_ai_cache_bypasses_total — counters sharing the llm_* label set (route_id, service_id, consumer, node, request_type, request_llm_model, llm_model)
• apisix_ai_cache_embedding_latency — histogram of embedding-call latency in milliseconds (the issue sketched _seconds, but the exporter's latency histograms use milliseconds with DEFAULT_BUCKETS, so this follows the house convention)

Semantics: Redis fail-open lookups count as MISS; fail_mode: error rejections record nothing; recording happens in log phase and is a silent no-op when the prometheus plugin is disabled. Docs: label sections added to prometheus.md (en + zh).

Which issue(s) this PR fixes:

Fixes #13290

Checklist

• I have explained the need for this PR and the problem it solves
• I have explained the changes or the new features added to this PR
• I have added tests corresponding to this change
• I have updated the documentation to reflect this change
• I have verified that this change is backward compatible (If not, please discuss on the APISIX mailing list first)