perf(energy): add 60s TTL cache for all 4 endpoints

Hydrogen overview was 1.8-2.0s (3 full-table aggregations on 66K rows).
With cache: cold 1 user/min eats the full query, all subsequent within
60s window return in ~10ms.

Implementation:
- New cache.ts with cached(key, loader) helper
- Per-key in-flight de-duplication: concurrent requests share one loader
- Each handler wrapped, cache key includes query params
  (e.g. "hydrogen/daily?range=last30&customer=external")
- TTL 60s as requested

Measured speedups:
- hydrogen/overview: 1.96s → 12ms (165x)
- hydrogen/daily:    270ms → 11ms (24x)
- electric/overview:  93ms →  9ms (10x)
- electric/monthly:   36ms →  9ms (4x)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/server/routes/energy/cache.ts

@@ -0,0 +1,44 @@
+/**
+ * 简单 TTL 内存缓存。
+ * 命中：直接返回缓存值；过期或未命中：运行 loader、存入缓存。
+ * 同一 key 并发请求只会触发一次 loader（共享 in-flight Promise）。
+ */
+
+interface Entry<T> {
+  value: T;
+  expiresAt: number;
+}
+
+const TTL_MS = 60 * 1000;
+
+const cache = new Map<string, Entry<unknown>>();
+const inflight = new Map<string, Promise<unknown>>();
+
+export async function cached<T>(key: string, loader: () => Promise<T>): Promise<T> {
+  const now = Date.now();
+  const hit = cache.get(key);
+  if (hit && hit.expiresAt > now) {
+    return hit.value as T;
+  }
+
+  // 同一 key 并发只跑一次 loader
+  const ongoing = inflight.get(key) as Promise<T> | undefined;
+  if (ongoing) return ongoing;
+
+  const p = loader()
+    .then(value => {
+      cache.set(key, { value, expiresAt: Date.now() + TTL_MS });
+      return value;
+    })
+    .finally(() => {
+      inflight.delete(key);
+    });
+  inflight.set(key, p as Promise<unknown>);
+  return p;
+}
+
+/** 仅用于测试或调试：清空所有缓存 */
+export function _clearEnergyCache() {
+  cache.clear();
+  inflight.clear();
+}