feat(llm): add configurable request/token rate limiting for LLM client

Introduce process-local and Redis-backed global rate limiting for the
shared LLM client, plus a dedicated generation consumer prefetch knob.

- Add request (RPM) and token (TPM) rate limits with burst controls,
  process or global scope, and fail-closed/process-degraded failover
- Wire the LLM client through NewWithRedis so limits can be shared
  across replicas via a Redis token bucket
- Add generation_consumer_prefetch to tune RabbitMQ generation
  consumers independently of the default prefetch
- Bump generation worker_concurrency to 24 and promote golang.org/x/time
  to a direct dependency
- Drop t.Parallel() from a few middleware/bootstrap tests

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

server/configs/config.local.yaml.example

@@ -70,6 +70,7 @@ rabbitmq:
   compliance_review_dlq: compliance.review.run.dlq
   compliance_review_dlq_routing_key: compliance.review.run.dlq
   consumer_prefetch: 10
+  generation_consumer_prefetch: 1
   publish_channel_pool_size: 32
   # Prefer environment override when needed:
   # export RABBITMQ_URL=amqp://geo:geo_dev@localhost:5672/geo
@@ -198,6 +199,16 @@ llm:
   temperature: 0.7
   reasoning_effort: minimal
   web_search_limit: 3
+  # Global Redis-backed budget shared by all replicas.
+  rate_limit_scope: global
+  rate_limit_key: ark:doubao-seed-2-0
+  rate_limit_failover_mode: process_degraded
+  rate_limit_failover_replicas: 10
+  request_rate_limit_rpm: 30000
+  request_rate_limit_burst: 3000
+  token_rate_limit_tpm: 5000000
+  token_rate_limit_burst: 500000
+  estimated_chars_per_token: 1
   # top_p: 0.95
   # Prefer environment override for secrets:
   # export LLM_API_KEY=your-api-key
@@ -233,7 +244,7 @@ retrieval:
 
 generation:
   queue_size: 128
-  worker_concurrency: 8
+  worker_concurrency: 24
   stream_enabled: false
   article_timeout: 8m
 

server/configs/config.yaml

@@ -83,6 +83,7 @@ rabbitmq:
   compliance_review_dlq: compliance.review.run.dlq
   compliance_review_dlq_routing_key: compliance.review.run.dlq
   consumer_prefetch: 10
+  generation_consumer_prefetch: 1
   publish_channel_pool_size: 32
 
 scheduler:
@@ -204,6 +205,16 @@ llm:
   temperature: 0.7
   reasoning_effort: minimal
   web_search_limit: 3
+  # Global Redis-backed budget shared by all replicas.
+  rate_limit_scope: global
+  rate_limit_key: ark:doubao-seed-2-0
+  rate_limit_failover_mode: process_degraded
+  rate_limit_failover_replicas: 10
+  request_rate_limit_rpm: 30000
+  request_rate_limit_burst: 3000
+  token_rate_limit_tpm: 5000000
+  token_rate_limit_burst: 500000
+  estimated_chars_per_token: 1
 
 compliance:
   enabled: true
@@ -232,7 +243,7 @@ retrieval:
 
 generation:
   queue_size: 128
-  worker_concurrency: 8
+  worker_concurrency: 24
   stream_enabled: false
   article_timeout: 8m
   task_lease_ttl: 10m