fix(machine-sync): backfill missing resource_id from provider state

sync_machines_with_provider only updated fields that already existed in
the update block. If a machine was stored before the handler fix (or via
a code path that omitted resource_id) it would never have resource_id
set, keeping it invisible to group_by_resource on the return path.

Add resource_id to both the needs_update predicate and the update block:
only backfill when provider knows the resource_id and the existing
record is missing it, leaving all other resource_id values untouched.

Author: fgogolli

src/orb/application/services/machine_sync_service.py

@@ -261,6 +261,7 @@ async def sync_machines_with_provider(
                         or existing.vpc_id != provider_machine.vpc_id
                         or existing.status_reason != provider_machine.status_reason
                         or existing.provider_data != provider_machine.provider_data
+                        or (provider_machine.resource_id and not existing.resource_id)
                     )
 
                     # Debug logging
@@ -288,6 +289,8 @@ async def sync_machines_with_provider(
                         machine_data["vpc_id"] = provider_machine.vpc_id
                         machine_data["version"] = existing.version + 1
                         machine_data["tags"] = provider_machine.tags
+                        if provider_machine.resource_id and not existing.resource_id:
+                            machine_data["resource_id"] = provider_machine.resource_id
 
                         updated_machine = Machine.model_validate(machine_data)
                         to_upsert.append(updated_machine)