Improve backend fallback logic for unknown queryId handling

[talh3747]

Problem

In RoutingManager.findBackendForUnknownQueryId(), after checking all backends by issuing HEAD requests, the fallback logic is hardcoded to use the first active backend with routing_group = 'adhoc':

gatewayBackendManager.getActiveAdhocBackends().get(0).getProxyTo()

This breaks in trino gateway environments where no backend is designated with routing_group = 'adhoc', because:

• getActiveAdhocBackends() executes:

  SELECT * FROM gateway_backend WHERE active = true AND routing_group = 'adhoc'

• If no row matches, the results list is empty → .get(0) throws IndexOutOfBoundsException, causing a 500 Internal Server Error for /v1/query/{nonexistent-id} requests.

Proposed fix

Modify the fallback logic to be more robust and generic:

• After failing to locate a backend by query ID, select fallback from any active backend, not just adhoc.

This ensures:

• No crash even if no adhoc routing_group exists
• Consistent behavior across environments

[Chaho12]

Would this resolve your issue?
#687