feat(compute): 算力中心 余额门控+充值+真实来源数值

- relay 匿名不再回落admin令牌(root无限额)→401或受限RELAY令牌(修无余额可用)
- /admin/compute/user-balance(充值add/subtract/override)+审计; compute_client.adjust_user_quota
- proxy follow_redirects(修gin尾斜杠301/307空数据); ping 返 version+quota_per_unit

app/services/compute_client.py

@@ -181,6 +181,22 @@ async def delete_token(token_id: int) -> dict:
     )
 
 
+async def adjust_user_quota(engine_user_id: int, value: int, mode: str = "add") -> dict:
+    """调整引擎用户余额/额度（充值/扣减/覆盖）。
+
+    对应引擎 ``POST /api/user/manage``，action="add_quota"：
+    - mode="add"       充值（增加余额）
+    - mode="subtract"  扣减
+    - mode="override"  直接设为该值
+    ``value`` 为配额单位（见 /api/status 的 quota_per_unit：500000 = $1）。
+    被充值用户的其他 PAT/会话会实时扣除，额度耗尽时引擎返回 403。
+    """
+    return await _request(
+        method="POST", path="/api/user/manage", headers=_admin_headers(),
+        json={"id": int(engine_user_id), "action": "add_quota", "value": int(value), "mode": mode},
+    )
+
+
 async def sync_user_enabled(username: str, enabled: bool) -> dict:
     """按 username 同步引擎用户启用/禁用（平台用户生命周期自动同步）。"""
     return await _request(
@@ -207,7 +223,7 @@ async def proxy(method: str, path: str, *, json_body: dict | None = None, params
         return {"status": 503, "body": '{"success": false, "message": "算力引擎内部令牌未配置（PINEAGENTS_INTERNAL_TOKEN）"}'}
     url = f"{_base()}/api{path}"
     try:
-        async with httpx.AsyncClient(timeout=config.COMPUTE_TIMEOUT) as client:
+        async with httpx.AsyncClient(timeout=config.COMPUTE_TIMEOUT, follow_redirects=True) as client:
             resp = await client.request(
                 method, url, json=json_body, params=params, headers=_admin_headers(),
             )