perf(queue): 按策略动态扩缩异步 Worker

将 River 执行容量按平台模型与用户组的有效并发策略动态调整，并为长任务续租、并发租约原子抢占和限流退避补充保护。\n\n统一平台模型限流继承语义，兼容历史 platformLimits/modelLimits，并为三个迁移图像模型建立独立并发租约。补充管理端显式继承/覆盖配置、指标、单元测试及隔离 PostgreSQL 验收。\n\n验证：go test ./... -count=1；pnpm lint；pnpm test；pnpm build；隔离 PostgreSQL 并发原子性/续租测试；128 任务与三分钟长任务动态 Worker 验收。

apps/api/internal/store/candidates.go

@@ -25,7 +25,7 @@ func (s *Store) ListModelCandidates(ctx context.Context, model string, modelType
        COALESCE(p.base_url, ''),
        p.auth_type, p.credentials, p.config, p.default_pricing_mode,
        p.default_discount_factor::float8, COALESCE(p.pricing_rule_set_id::text, ''),
-       p.retry_policy, p.rate_limit_policy,
+       p.retry_policy, p.rate_limit_policy, COALESCE(b.default_rate_limit_policy, '{}'::jsonb),
        COALESCE(p.dynamic_priority, p.priority) AS effective_priority,
        m.id::text, COALESCE(m.base_model_id::text, ''), COALESCE(b.canonical_model_key, ''),
        COALESCE(NULLIF(m.provider_model_name, ''), m.model_name), m.model_name, COALESCE(m.model_alias, ''),
@@ -54,8 +54,11 @@ func (s *Store) ListModelCandidates(ctx context.Context, model string, modelType
        COALESCE(b.base_billing_config, '{}'::jsonb), m.billing_config, m.billing_config_override,
        m.pricing_mode, COALESCE(m.discount_factor, 0)::float8, COALESCE(m.pricing_rule_set_id::text, ''),
        COALESCE(b.pricing_rule_set_id::text, ''),
-       m.permission_config, m.retry_policy, m.rate_limit_policy, COALESCE(m.runtime_policy_set_id::text, COALESCE(b.runtime_policy_set_id::text, '')),
+       m.permission_config, m.retry_policy, m.rate_limit_policy, m.rate_limit_policy_mode,
+       COALESCE(m.runtime_policy_set_id::text, COALESCE(b.runtime_policy_set_id::text, '')),
+       (m.runtime_policy_set_id IS NOT NULL),
        COALESCE(NULLIF(m.runtime_policy_override, '{}'::jsonb), b.runtime_policy_override, '{}'::jsonb),
+       COALESCE(m.runtime_policy_override, '{}'::jsonb),
        COALESCE(rp.retry_policy, '{}'::jsonb), COALESCE(rp.rate_limit_policy, '{}'::jsonb),
        COALESCE(rp.auto_disable_policy, '{}'::jsonb), COALESCE(rp.degrade_policy, '{}'::jsonb),
        COALESCE(con.active, 0)::float8,
@@ -177,6 +180,7 @@ WHERE p.status = 'enabled'
 		var platformConfig []byte
 		var platformRetryPolicy []byte
 		var platformRateLimitPolicy []byte
+		var baseRateLimitPolicy []byte
 		var capabilities []byte
 		var capabilityOverride []byte
 		var baseBilling []byte
@@ -186,6 +190,7 @@ WHERE p.status = 'enabled'
 		var modelRetryPolicy []byte
 		var modelRateLimitPolicy []byte
 		var runtimePolicyOverride []byte
+		var rateLimitRuntimeOverride []byte
 		var runtimeRetryPolicy []byte
 		var runtimeRateLimitPolicy []byte
 		var autoDisablePolicy []byte
@@ -222,6 +227,7 @@ WHERE p.status = 'enabled'
 			&item.PlatformPricingRuleSetID,
 			&platformRetryPolicy,
 			&platformRateLimitPolicy,
+			&baseRateLimitPolicy,
 			&item.PlatformPriority,
 			&item.PlatformModelID,
 			&item.BaseModelID,
@@ -244,8 +250,11 @@ WHERE p.status = 'enabled'
 			&permissionConfig,
 			&modelRetryPolicy,
 			&modelRateLimitPolicy,
+			&item.ModelRateLimitPolicyMode,
 			&item.RuntimePolicySetID,
+			&item.RuntimePolicyExplicit,
 			&runtimePolicyOverride,
+			&rateLimitRuntimeOverride,
 			&runtimeRetryPolicy,
 			&runtimeRateLimitPolicy,
 			&autoDisablePolicy,
@@ -274,6 +283,7 @@ WHERE p.status = 'enabled'
 		item.PlatformConfig = decodeObject(platformConfig)
 		item.PlatformRetryPolicy = decodeObject(platformRetryPolicy)
 		item.PlatformRateLimitPolicy = decodeObject(platformRateLimitPolicy)
+		item.BaseRateLimitPolicy = decodeObject(baseRateLimitPolicy)
 		item.Capabilities = decodeObject(capabilities)
 		item.CapabilityOverride = decodeObject(capabilityOverride)
 		item.BaseBillingConfig = decodeObject(baseBilling)
@@ -283,6 +293,7 @@ WHERE p.status = 'enabled'
 		item.ModelRetryPolicy = decodeObject(modelRetryPolicy)
 		item.ModelRateLimitPolicy = decodeObject(modelRateLimitPolicy)
 		item.RuntimePolicyOverride = decodeObject(runtimePolicyOverride)
+		item.RateLimitRuntimeOverride = decodeObject(rateLimitRuntimeOverride)
 		item.RuntimeRetryPolicy = decodeObject(runtimeRetryPolicy)
 		item.RuntimeRateLimitPolicy = decodeObject(runtimeRateLimitPolicy)
 		item.AutoDisablePolicy = decodeObject(autoDisablePolicy)
@@ -303,7 +314,15 @@ WHERE p.status = 'enabled'
 			LastObservedUnix:  cacheLastObservedUnix,
 		})
 		applyRuntimeCandidateLoad(&item, runtimeCandidateLoadInput{
-			Policy:            effectiveModelRateLimitPolicy(item.PlatformRateLimitPolicy, item.RuntimeRateLimitPolicy, item.RuntimePolicySetID, item.RuntimePolicyOverride, item.ModelRateLimitPolicy),
+			Policy: EffectiveRateLimitPolicy(EffectiveRateLimitPolicyInput{
+				BasePolicy:            item.BaseRateLimitPolicy,
+				PlatformPolicy:        item.PlatformRateLimitPolicy,
+				RuntimePolicy:         item.RuntimeRateLimitPolicy,
+				RuntimePolicyExplicit: item.RuntimePolicyExplicit,
+				RuntimePolicyOverride: item.RateLimitRuntimeOverride,
+				ModelPolicy:           item.ModelRateLimitPolicy,
+				ModelPolicyMode:       item.ModelRateLimitPolicyMode,
+			}),
 			ConcurrentActive:  concurrentActive,
 			QueuedWaiting:     queuedWaiting,
 			RPMUsed:           rpmUsed,