[ML] Data Drift View bugs

[valeriy42]

Hi @qn895,

great job on implementing the Data Drift view. I gave it a spin and noticed a couple of issues I wanted to document here:

1. KS-Test value is missing when value ranges are disjunct. I used iris dataset and filtered two different classes for reference and comparison. Now petal_length doesn't show any Similarity p-value and "Drift detected" is noted as "No", although the values are completely different.

2. KS-Test p-value for pdays on the bank-marketing dataset is very small, although reference and comparison distributions are identical. I guess we have a numerical issue here we need to investigate.

3. I really like how intuitive the interval selection is. However, when I was playing with the rulers on the farequote dataset they went out of sync.

Those are the only bugs I was able to notice. Overall, I think it is a great feature and it will be very useful for operationalizing ML in the stack.

[elasticmachine]

Pinging @elastic/ml-ui (:ml)

[qn895]

As mentioned in Slack, issue 1 and 2 require more investigation into the ES queries we are currently using. For the third issue with the ruler, we have an enhancement slated for 8.12 to revamp the user experience completely.

[qn895]

Closing this issue as they have been fixed after #167911 and #168757