[Metrics UI] Fix sync issues with alert intervals and metrics explorer intervals

[Zacqary]

The values displayed in the Metrics Explorer and the values reported by an alert are close to each other but not the same. Note the data point on the left is 2.2, but the log message sent at roughly the same time reports a value of 1.8:

Is this as simple as making sure "Round to the minute" is turned on both on the Metrics Explorer and the alert executor? Or is it a more complex aggregation issue?

[elasticmachine]

Pinging @elastic/logs-metrics-ui (Team:logs-metrics-ui)

[Zacqary]

Zooming out to 2 hours of data syncs up the metrics explorer to one data point per minute, and these definitely start to look closer to the alert's interval. Maybe less of a bug in the executor and more of a UX issue?

How can we decrease confusion? Right now the user has to use trial and error to figure out how to display the same data point interval that their alert would be capturing.

Perhaps also forcing alerts to round to the nearest minute/hour/etc. would improve things.

[simianhacker]

It's where the bucket falls. The code for the alert executor re-aligns the buckets to the right side (bucketing starts from the to) of the chart where TSVB starts the bucketing from the left side of the chart (bucketing starts from the from).

[Zacqary]

Opened a PR to round alerts to the nearest unit, which seems to greatly improve accuracy relative to the Metrics Explorer. It's still not perfect, and @simianhacker and I discussed using the Snapshot API for the Metrics Explorer (and maybe threshold alerts too) instead of TSVB to improve accuracy.

@simianhacker if you want to open up a separate issue for that then I can set #71172 to close this issue