[ML] Dataframe Analytics: Update feature_influence format

[walterra]

Until now, the format for feature influence looked like this: Next to outlier_score there are feature_influence.<feature_name> entries.

This format is changing as having the feature name in the field name may become problematic (field explosion, invalid field chars, etc.)

The updated format will look like this:

{
  ...
  "results_field": {
    "outlier_score": 0.42,
    "feature_influence": [
      {
        "feature_name": "name1",
        "influence": 0.333
      },
      {
        "feature_name": "name2",
        "influence": 0.01
      }
    ]
  }
}

Since the feature is still experimental, we don't necessarily have to support the old format, still, it's worth discussing if we want to support it. We have to implement that the code recognizes a possible old format nonetheless, so we could just as well then support correctly showing the old data.

Elasticsearch related PR: elastic/elasticsearch#62068

---

• Fix cell colour coding [ML] DF Analytics: Collapsable sections on results pages #76641
• Display a callout for legacy data structures saying color coding is not available [ML] Data Frame Analytics: Fix race condition and support for feature influence legacy format. #81123

[elasticmachine]

Pinging @elastic/ml-ui (:ml)

[peteharverson]

After the updates were made for the new feature_influence format, I think we should aim for:

• new jobs display full results as currently, with color coded cells to indicte the influence of that feature
• old jobs - results page opens without error, with table populated with data, but without the color coding of cells, plus a callout to indicate that the results are from a legacy format job

[walterra]

The callout for legacy feature influence format was added in #81123, closing this issue.