[ML] File upload: enabling check for model allocations (elastic#197395)

Enables the previously commented out check for `num_allocations` when
listing the inference endpoints.

The adaptive allocation count can drop to 0, but it is still valid for
use. Uploading a file will cause it to be re-deployed.

Related to es PRs elastic/elasticsearch#115233
and elastic/elasticsearch#115095

Follow on from elastic#196577

(cherry picked from commit 66b2447)

x-pack/plugins/data_visualizer/server/routes.ts

@@ -96,9 +96,9 @@ export function routes(coreSetup: CoreSetup<StartDeps, unknown>, logger: Logger)
 
           const filteredInferenceEndpoints = endpoints.filter((endpoint) => {
             return (
-              endpoint.task_type === 'sparse_embedding' || endpoint.task_type === 'text_embedding'
-              // TODO: add this back in when the fix has made it into es in 8.16
-              // && endpoint.service_settings.num_allocations > 0
+              (endpoint.task_type === 'sparse_embedding' ||
+                endpoint.task_type === 'text_embedding') &&
+              endpoint.service_settings.num_allocations >= 0
             );
           });