feat(worker): 实现集群限流与自适应负载

保留平台模型 RPM、TPM 和并发策略语义，增加 PostgreSQL 集群级租约、饱和候选重选和多平台自动负载，避免突发任务固定等待首个平台。\n\n新增 Worker 实时负载采样、自适应 active/heavy 容量、心跳与管理端指标，并扩展本地 acceptance runner，覆盖三 Worker、同模型三平台 2/4/6 并发和 48 个带图视频突发任务。\n\n验证：go test ./...、go vet ./...、PostgreSQL 跨 Store 集成测试、gofmt、bash -n、ShellCheck 及本地集群 provider-burst 验收通过；48/48 成功，无越限、重复提交、重复计费、重复回调或终态资源泄漏。

apps/api/internal/config/config.go

@@ -79,6 +79,7 @@ type Config struct {
 	AsyncWorkerHardLimit                          int
 	AsyncWorkerInstanceHardLimit                  int
 	AsyncWorkerRefreshIntervalSeconds             int
+	AsyncWorkerLoadMode                           string
 	AsyncAdmissionMicrobatchSize                  int
 	AsyncAdmissionDispatcherEnabled               bool
 	AsyncAdmissionDispatcherConfigured            bool
@@ -189,6 +190,7 @@ func Load() Config {
 		),
 		AsyncWorkerInstanceHardLimit:      envIntValidated("AI_GATEWAY_ASYNC_WORKER_INSTANCE_HARD_LIMIT", 32),
 		AsyncWorkerRefreshIntervalSeconds: envIntValidated("AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS", 5),
+		AsyncWorkerLoadMode:               strings.ToLower(strings.TrimSpace(env("AI_GATEWAY_WORKER_LOAD_MODE", "adaptive"))),
 		AsyncAdmissionMicrobatchSize:      envIntValidated("AI_GATEWAY_ASYNC_ADMISSION_MICROBATCH_SIZE", 8),
 		AsyncAdmissionDispatcherEnabled:   envValue("AI_GATEWAY_ASYNC_ADMISSION_DISPATCHER_ENABLED") == "true",
 		AsyncAdmissionDispatcherConfigured: envValue(
@@ -299,6 +301,11 @@ func (c Config) Validate() error {
 	if c.AsyncWorkerRefreshIntervalSeconds < 1 {
 		return errors.New("AI_GATEWAY_ASYNC_WORKER_REFRESH_INTERVAL_SECONDS must be positive")
 	}
+	switch strings.ToLower(strings.TrimSpace(c.AsyncWorkerLoadMode)) {
+	case "", "adaptive", "legacy":
+	default:
+		return errors.New("AI_GATEWAY_WORKER_LOAD_MODE must be adaptive or legacy")
+	}
 	if c.AsyncAdmissionMicrobatchSize < 1 || c.AsyncAdmissionMicrobatchSize > 32 {
 		return errors.New("AI_GATEWAY_ASYNC_ADMISSION_MICROBATCH_SIZE must be between 1 and 32")
 	}