[Lens] Sparse histogram data leads to overly wide bars in xy chart

[flash1293]

Kibana version: master

Describe the bug: If the "intervals" function is used, completely empty buckets are dropped and not part of the table passed to the visualization. For sparse data, this causes a problem because the used interval can't be inferred from the data, leading to elastic-charts guessing. This is very similar to #81580 (for date histograms)

Steps to reproduce:

1. Go to Lens with logs sample data with a small time range
2. Use intervals of bytes on the x axis and count on the y axis
3. Observe the bars to be much wider than the underlying interval used

In both of these, the used interval is 100, but it looks like 900

Expected behavior:

Bars should show the correct width of 100

Any additional context:

There are a few options what to do:

• Implement a solution to pass the used interval as part of the meta data similar to the date histogram. I recommend this solution even if it's a bit more complex because it will allow us to leverage this information in other places as well (like in the config UI) and right now there's no way to determine the currently used interval
• Just leave it as it is right now - the chart is not technically wrong, it could just provide higher granularity information
• Fetch empty buckets as well so elastic-charts can infer the actual interval from the empty buckets

cc @ppisljar what do you think about option 1 here? This case is very similar to the date histogram "auto" interval used, so IMHO it would make sense to use the same solution.

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[ppisljar]

i agree option 1 makes sense