[ML] Reenable Semantic Search yaml tests (#90956)

The search hits cannot be asserted on as the simple model used 
in the test relies on randomisation and produces different results 
on different platforms. The test is covered elsewhere so reenable
without the assertions here.

x-pack/plugin/src/yamlRestTest/resources/rest-api-spec/test/ml/semantic_search.yml

@@ -1,5 +1,19 @@
 # This test uses the simple text embedding model from
-# SemanticSearchIT to create the dense vectors
+# SemanticSearchIT.java to create the dense vectors. The
+# model creates random tensors seeded by the hash of
+# input string, we expect the same input to produce
+# the same output. However, not all platforms produce
+# the same random tensor for the same seed. The model
+# creates different tensors on different platforms and
+# OS's, therefore the top search hits will be different
+# on different platforms meaning the test cannot assert
+# on the top hits. SemanticSearchIT does not suffer
+# from this limitation as it dynamically creates the
+# embeddings and documents rather than using hard-coded
+# values. It is important to keep this test for the
+# language clients to snoop the response format but the
+# assertions can be skipped here because of the test
+# coverage in SemanticSearchIT
 
 setup:
   - skip:
@@ -110,10 +124,6 @@ setup:
 
 ---
 "Test semantic search":
-  - skip:
-      reason: "https://github.com/elastic/elasticsearch/issues/90904"
-      version: "8.6.0 - "
-
   - do:
       ml.start_trained_model_deployment:
         model_id: text_embedding_model
@@ -132,7 +142,8 @@ setup:
             num_candidates: 10
   - gte: { inference_took: 0 }
   - match: { hits.total.value: 3 }
-  - match: { hits.hits.0._source.source_text: "the octopus comforter smells" }
+    # See comment at the top of the file for the reason why
+    # the test cannot match on the text field
 
 ---
 "Knn field is not a vector":