/active_series: improve json encoding performance

[flxbk]

What this PR does

This adds a benchmark for the querier's ActiveSeriesCardinalityHandler and uses json-iterator to replace the standard library encoding/json to improve marshaling performance.

Here are benchmark results comparing the first two commits in this PR (std vs json-iterator encoding).

goos: darwin
goarch: arm64
pkg: github.com/grafana/mimir/pkg/querier
                                 │   stdjson    │              jsoniter              │
                                 │    sec/op    │   sec/op     vs base               │
ActiveSeriesHandler_ServeHTTP-10   987.62µ ± 0%   98.82µ ± 0%  -89.99% (p=0.002 n=6)

                                 │   stdjson    │              jsoniter               │
                                 │     B/op     │     B/op      vs base               │
ActiveSeriesHandler_ServeHTTP-10   965.7Ki ± 0%   148.8Ki ± 0%  -84.59% (p=0.002 n=6)

                                 │   stdjson    │             jsoniter              │
                                 │  allocs/op   │ allocs/op   vs base               │
ActiveSeriesHandler_ServeHTTP-10   13139.0 ± 0%   132.0 ± 0%  -99.00% (p=0.002 n=6)

Which issue(s) this PR fixes or relates to

n/a

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[charleskorn]

LGTM

[Logiraptor]

Should this have a Status: "success" field set as well?

[flxbk]

Prometheus does set that field, however it's not quite clear to me why. Isn't the HTTP status code a good enough indication that the payload is going to be ok/parseable as expected?

[Logiraptor]

Sure, I don't think we need the field either, but I think right now it is returning Status: "", which is not useful. So I'd say we either keep it and conform to the prometheus response (which includes setting the error fields), or just return Data and nothing else.

[flxbk]

Ok, makes sense, I'll stick with the custom struct then.