refactor(saas): 架构重构 + 性能优化 — 借鉴 loco-rs 模式

Phase 0: 知识库
- docs/knowledge-base/loco-rs-patterns.md — loco-rs 10 个可借鉴模式研究

Phase 1: 数据层重构
- crates/zclaw-saas/src/models/ — 15 个 FromRow 类型化模型
- Login 3 次查询合并为 1 次 AccountLoginRow 查询
- 所有 service 文件从元组解构迁移到 FromRow 结构体

Phase 2: Worker + Scheduler 系统
- crates/zclaw-saas/src/workers/ — Worker trait + 5 个具体实现
- crates/zclaw-saas/src/scheduler.rs — TOML 声明式调度器
- crates/zclaw-saas/src/tasks/ — CLI 任务系统

Phase 3: 性能修复
- Relay N+1 查询 → 精准 SQL (relay/handlers.rs)
- Config RwLock → AtomicU32 无锁 rate limit (state.rs, middleware.rs)
- SSE std::sync::Mutex → tokio::sync::Mutex (relay/service.rs)
- /auth/refresh 阻塞清理 → Scheduler 定期执行

Phase 4: 多环境配置
- config/saas-{development,production,test}.toml
- ZCLAW_ENV 环境选择 + ZCLAW_SAAS_CONFIG 精确覆盖
- scheduler 配置集成到 TOML

crates/zclaw-saas/src/relay/key_pool.rs

@@ -4,6 +4,7 @@
 
 use sqlx::PgPool;
 use crate::error::{SaasError, SaasResult};
+use crate::models::{ProviderKeySelectRow, ProviderKeyRow};
 use crate::crypto;
 
 /// 解密 key_value (如果已加密)，否则原样返回
@@ -40,7 +41,7 @@ pub async fn select_best_key(db: &PgPool, provider_id: &str, enc_key: &[u8; 32])
     let current_minute = chrono::Utc::now().format("%Y-%m-%dT%H:%M").to_string();
 
     // 获取所有活跃 Key
-    let rows: Vec<(String, String, i32, Option<i64>, Option<i64>, Option<String>)> =
+    let rows: Vec<ProviderKeySelectRow> =
         sqlx::query_as(
             "SELECT id, key_value, priority, max_rpm, max_tpm, quota_reset_interval
              FROM provider_keys
@@ -89,18 +90,18 @@ pub async fn select_best_key(db: &PgPool, provider_id: &str, enc_key: &[u8; 32])
     }
 
     // 检查滑动窗口使用量
-    for (id, key_value, priority, max_rpm, max_tpm, quota_reset_interval) in rows {
+    for row in rows {
         // 检查 RPM 限额
-        if let Some(rpm_limit) = max_rpm {
+        if let Some(rpm_limit) = row.max_rpm {
             if rpm_limit > 0 {
                 let window: Option<(i64,)> = sqlx::query_as(
                     "SELECT COALESCE(SUM(request_count), 0) FROM key_usage_window
                      WHERE key_id = $1 AND window_minute = $2"
-                ).bind(&id).bind(&current_minute).fetch_optional(db).await?;
+                ).bind(&row.id).bind(&current_minute).fetch_optional(db).await?;
 
                 if let Some((count,)) = window {
                     if count >= rpm_limit {
-                        tracing::debug!("Key {} hit RPM limit ({}/{})", id, count, rpm_limit);
+                        tracing::debug!("Key {} hit RPM limit ({}/{})", row.id, count, rpm_limit);
                         continue;
                     }
                 }
@@ -108,16 +109,16 @@ pub async fn select_best_key(db: &PgPool, provider_id: &str, enc_key: &[u8; 32])
         }
 
         // 检查 TPM 限额
-        if let Some(tpm_limit) = max_tpm {
+        if let Some(tpm_limit) = row.max_tpm {
             if tpm_limit > 0 {
                 let window: Option<(i64,)> = sqlx::query_as(
                     "SELECT COALESCE(SUM(token_count), 0) FROM key_usage_window
                      WHERE key_id = $1 AND window_minute = $2"
-                ).bind(&id).bind(&current_minute).fetch_optional(db).await?;
+                ).bind(&row.id).bind(&current_minute).fetch_optional(db).await?;
 
                 if let Some((tokens,)) = window {
                     if tokens >= tpm_limit {
-                        tracing::debug!("Key {} hit TPM limit ({}/{})", id, tokens, tpm_limit);
+                        tracing::debug!("Key {} hit TPM limit ({}/{})", row.id, tokens, tpm_limit);
                         continue;
                     }
                 }
@@ -125,17 +126,17 @@ pub async fn select_best_key(db: &PgPool, provider_id: &str, enc_key: &[u8; 32])
         }
 
         // 此 Key 可用 — 解密 key_value
-        let decrypted_kv = decrypt_key_value(&key_value, enc_key)?;
+        let decrypted_kv = decrypt_key_value(&row.key_value, enc_key)?;
         return Ok(KeySelection {
             key: PoolKey {
-                id: id.clone(),
+                id: row.id.clone(),
                 key_value: decrypted_kv,
-                priority,
-                max_rpm,
-                max_tpm,
-                quota_reset_interval,
+                priority: row.priority,
+                max_rpm: row.max_rpm,
+                max_tpm: row.max_tpm,
+                quota_reset_interval: row.quota_reset_interval,
             },
-            key_id: id,
+            key_id: row.id,
         });
     }
 
@@ -229,7 +230,7 @@ pub async fn list_provider_keys(
     db: &PgPool,
     provider_id: &str,
 ) -> SaasResult<Vec<serde_json::Value>> {
-    let rows: Vec<(String, String, String, i32, Option<i64>, Option<i64>, Option<String>, bool, Option<String>, Option<String>, i64, i64, String, String)> =
+    let rows: Vec<ProviderKeyRow> =
         sqlx::query_as(
             "SELECT id, provider_id, key_label, priority, max_rpm, max_tpm, quota_reset_interval, is_active,
                    last_429_at, cooldown_until, total_requests, total_tokens, created_at, updated_at
@@ -238,20 +239,20 @@ pub async fn list_provider_keys(
 
     Ok(rows.into_iter().map(|r| {
         serde_json::json!({
-            "id": r.0,
-            "provider_id": r.1,
-            "key_label": r.2,
-            "priority": r.3,
-            "max_rpm": r.4,
-            "max_tpm": r.5,
-            "quota_reset_interval": r.6,
-            "is_active": r.7,
-            "last_429_at": r.8,
-            "cooldown_until": r.9,
-            "total_requests": r.10,
-            "total_tokens": r.11,
-            "created_at": r.12,
-            "updated_at": r.13,
+            "id": r.id,
+            "provider_id": r.provider_id,
+            "key_label": r.key_label,
+            "priority": r.priority,
+            "max_rpm": r.max_rpm,
+            "max_tpm": r.max_tpm,
+            "quota_reset_interval": r.quota_reset_interval,
+            "is_active": r.is_active,
+            "last_429_at": r.last_429_at,
+            "cooldown_until": r.cooldown_until,
+            "total_requests": r.total_requests,
+            "total_tokens": r.total_tokens,
+            "created_at": r.created_at,
+            "updated_at": r.updated_at,
         })
     }).collect())
 }