Pool zstd encoding/decoding

[zalegrala]

What this PR does:

Here we pool the zstd encoders and decoders for better performance when handling the tenant index.

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 zstdPooling.txt
goos: linux
goarch: amd64
pkg: github.com/grafana/tempo/tempodb/backend
cpu: AMD Ryzen 7 PRO 7840U w/ Radeon 780M Graphics
                    │   main.txt   │           zstdPooling.txt           │
                    │    sec/op    │   sec/op     vs base                │
IndexUnmarshalPb-16   20.645µ ± 1%   3.694µ ± 2%  -82.11% (p=0.000 n=10)

                    │    main.txt    │           zstdPooling.txt            │
                    │      B/op      │     B/op      vs base                │
IndexUnmarshalPb-16   154.434Ki ± 0%   3.709Ki ± 1%  -97.60% (p=0.000 n=10)

                    │  main.txt   │          zstdPooling.txt           │
                    │  allocs/op  │ allocs/op   vs base                │
IndexUnmarshalPb-16   100.00 ± 0%   59.00 ± 0%  -41.00% (p=0.000 n=10)

[joe-elliott]

changelog!

amazing benches. should be a good one

[javiermolinar]

awesome!