[ML] Incorrect connecting of points in anomaly charts when data is sparse

[peteharverson]

Chart points in the Anomaly Explorer charts can be incorrectly connected when the data is sparse and there are gaps in the data.

Example job config on gallery data set:

"analysis_config": {
    "bucket_span": "6h",
    "detectors": [
      {
        "detector_description": "sum(bytes) partitionfield=clientip",
        "function": "sum",
        "field_name": "bytes",
        "partition_field_name": "clientip",
        "detector_index": 0
      }
    ],
    "influencers": [
      "clientip"
    ],
    "model_prune_window": "30d"
  },

Anomaly Explorer charts:

Compared to chart in the Single Metric Viewer where the connections are as intended, with connection to points at y axis value of 0 and gaps in data:

[elasticmachine]

Pinging @elastic/ml-ui (:ml)