[Bugfix][Router] Return 503 when failover ends on a backend client error

[Mazyod]

FIX #1163

Today, when failover ends before response headers and the last error is an aiohttp.ClientError (connection refused, DNS failure, and similar), route_general_request re-raises it. Clients get a plain-text 500, and uvicorn logs Exception in ASGI application with a traceback.

This PR returns the same 503 status and JSON body as proxy_multipart_request, {"error": "Failed to connect to backend: ..."}, and adds X-Request-Id to the response. Other exceptions are re-raised as before, mid-stream failures are unchanged, and the existing failover tests are untouched.

Testing:

• New unit test test_unreachable_backends_return_503_after_all_attempts in src/tests/test_instance_failover.py. It fails on main and passes with this change.
• Against a closed port, the router returned 503 application/json with X-Request-Id and logged no ASGI traceback. Main returns 500 text/plain and logs a traceback.
• All 273 router tests pass on Python 3.12, and pre-commit and ruff check pass.

---

• Make sure the code changes pass the pre-commit checks.
• Sign-off your commit by using -s when doing git commit
• Try to classify PRs for easy understanding of the type of changes, such as [Bugfix], [Feat], and [CI].

[gemini-code-assist]

Code Review

This pull request updates the route_general_request function in src/vllm_router/services/request_service/request.py to handle aiohttp.ClientError when backends are unreachable, returning a 503 JSONResponse with the request ID and a descriptive error message. Additionally, a new unit test test_unreachable_backends_return_503_after_all_attempts has been added to src/tests/test_instance_failover.py to verify this error handling behavior. There are no review comments to address, and I have no further feedback to provide.