server: Add cached_tokens info to oaicompat responses (#19361)

* tests : fix fetch_server_test_models.py

* server: to_json_oaicompat cached_tokens

Adds OpenAI and Anthropic compatible information about the
number of cached prompt tokens used in a response.

scripts/fetch_server_test_models.py

@@ -95,9 +95,9 @@ if __name__ == '__main__':
             '-p', 'Hey',
             '--no-warmup',
             '--log-disable',
-            '-no-cnv']
+            '-st']
         if m.hf_file != 'tinyllamas/stories260K.gguf' and 'Mistral-Nemo' not in m.hf_repo:
-            cmd.append('-fa')
+            cmd += ('-fa', 'on')
         try:
             subprocess.check_call(cmd)
         except subprocess.CalledProcessError: