[ML] Continuous data frames should show more informative progress

[sophiec20]

Found in 7.3.0-BC1

Continuous data frames have a naive progress animation to show that the persistent task is started. However, this is purely an animation and it is difficult to tell how much progress is being made.

Due to the nature of the implementation, there is not a perfect way to display progress. The most useful indicator field we have is the time_upper_bound_millis of the last completed checkpoint.

This is available from GET _data_frame/transforms/blah/_stats (also takes ?human=true for read-able date)

This is the last time that the checkpoint was applied minus the query delay. This value will be updated upon completion of the next checkpoint.

It would be good to display this value on the transform list, without having to use the row expansion.

Whilst checkpoint 1 is underway (checkpoint 0) this value will probably not exist.

For batch data frames, once they have completed, this value will be available for checkpoint 1.

"checkpointing" : {
        "current" : {
          "timestamp" : "2019-07-04T17:28:45.845Z",
          "timestamp_millis" : 1562261325845,
          "time_upper_bound" : "2019-07-04T17:28:44.845Z",
          "time_upper_bound_millis" : 1562261324845
...

Note that we are planning to change "current" to "last".

Update: this value is currently only returned if the indexer is indexing .. we should not attempt to display it unless it has more stability. Awaits backend change.

[elasticmachine]

Pinging @elastic/ml-ui