MQE: Add support for delta function

[lamida]

What this PR does

Which issue(s) this PR fixes or relates to

Fixes #

Checklist

• Tests updated.
• na Documentation added.
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX].
• na about-versioning.md updated with experimental features.

[charleskorn]

Could you please add some tests for delta over native histograms?

It'd be good include native histograms with custom buckets, as well as scenarios that trigger the various annotations. (Have a look at our tests for rate and increase for some inspiration, and look at TestAnnotations in engine_test.go.)

[charleskorn]

Could you please also add some tests for the case where the value (float or histogram) increases and decreases during the range, to ensure that we're correctly ignoring resets?

[charleskorn]

I'm not sure this is quite right - the docs for delta say:

delta(v range-vector) calculates the difference between the first and last value of each time series element in a range vector v, returning an instant vector with the given deltas and equivalent labels. The delta is extrapolated to cover the full time range as specified in the range vector selector...

I see two issues with the implementation here:

• we're not doing extrapolation
• we're considering resets in delta, but we want to consider just the first and last value

I wonder if it would be clearer and simpler for delta to not use rate above, and instead be implemented separately and call calculateFloatRate and calculateHistogramRate to perform extrapolation.

[lamida]

Actually I was looking at how prometheus engine was implementing delta. But let me check at this again and try to apply your suggestion.