Native histogram performance

[zalegrala]

What this PR does:

Improve performance in the native histograms implementations. This is following in the footsteps of #4232.

Which issue(s) this PR fixes:
Fixes #

Checklist

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

[zalegrala]

❯ benchstat main.txt wip.txt
goos: linux
goarch: amd64
pkg: github.com/grafana/tempo/modules/generator
cpu: AMD Ryzen 7 PRO 7840U w/ Radeon 780M Graphics
           │   main.txt   │               wip.txt               │
           │    sec/op    │   sec/op     vs base                │
Collect-16   4.960µ ± 10%   4.421µ ± 5%  -10.86% (p=0.002 n=10)

           │  main.txt   │              wip.txt               │
           │ heap_in_use │ heap_in_use  vs base               │
Collect-16   17.81M ± 1%   17.42M ± 1%  -2.16% (p=0.000 n=10)

           │   main.txt   │               wip.txt               │
           │     B/op     │     B/op      vs base               │
Collect-16   14.22Ki ± 3%   13.59Ki ± 3%  -4.44% (p=0.000 n=10)

           │  main.txt   │            wip.txt            │
           │  allocs/op  │  allocs/op   vs base          │
Collect-16   40.00 ± 40%   31.00 ± 35%  ~ (p=0.137 n=10)

I'm seeing some randomness in the benchmarks, which makes is less reliable for comparison.

[joe-elliott]

some Qs but i think it looks great

[joe-elliott]

Let's go!