[ML] Add hint about using input_output in the inference processor to error message

[davidkyle]

If using the Inference Ingest processor the input_output option must be used instead of field_mappings . The error message does not make it clear to the user what the problem is.

For example:

# Create an ingest pipeline with inference on an inference endpoint and field_map
PUT _ingest/pipeline/elser_pipeline
{
  "description": "Pipeline to run ELSER model on an inference endpoint",
  "processors": [
    {
      "inference": {
        "model_id": "my_elser",
        "inference_config": {
          "text_expansion": {
            "results_field": "text_embedding"
          }
        },
        "field_map": {
          "text": "text"
        }
      }
    }
  ]
}

# When using the pipeline it errors with
{
  "error": {
    "root_cause": [
      {
        "type": "status_exception",
        "reason": "[my_elser] is configured for the _inference API and does not accept documents as input"
      }
    ],
    "type": "status_exception",
    "reason": "[my_elser] is configured for the _inference API and does not accept documents as input"
  },
  "status": 400
}

The limitation is documented at https://www.elastic.co/docs/reference/enrich-processor/inference-processor

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)