Optimize the decoding of SampleStream

[alanprot]

What this PR does:
Create jsoniter SampleStream Decoder to optimize its deserialization.

QueryFrontend can benefit from this as it deserialize the responses from the queriers:

goos: linux
goarch: amd64
pkg: github.com/cortexproject/cortex/pkg/querier/tripperware/instantquery
cpu: Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz
                           │   /tmp/old   │              /tmp/new               │
                           │    sec/op    │    sec/op     vs base               │
_Decode/100_samples-32       280.5µ ± ∞ ¹   209.3µ ± ∞ ¹  -25.36% (p=0.008 n=5)
_Decode/1000_samples-32      2.875m ± ∞ ¹   2.133m ± ∞ ¹  -25.82% (p=0.008 n=5)
_Decode/10000_samples-32     28.83m ± ∞ ¹   21.55m ± ∞ ¹  -25.24% (p=0.008 n=5)
_Decode/100000_samples-32    304.5m ± ∞ ¹   247.3m ± ∞ ¹  -18.79% (p=0.008 n=5)
_Decode/1000000_samples-32    3.060 ± ∞ ¹    2.379 ± ∞ ¹  -22.27% (p=0.008 n=5)
geomean                      29.32m         22.42m        -23.54%
¹ need >= 6 samples for confidence interval at level 0.95

                           │   /tmp/old    │               /tmp/new               │
                           │     B/op      │     B/op       vs base               │
_Decode/100_samples-32       129.3Ki ± ∞ ¹   109.1Ki ± ∞ ¹  -15.67% (p=0.008 n=5)
_Decode/1000_samples-32      1.213Mi ± ∞ ¹   1.001Mi ± ∞ ¹  -17.49% (p=0.008 n=5)
_Decode/10000_samples-32     15.64Mi ± ∞ ¹   13.36Mi ± ∞ ¹  -14.53% (p=0.008 n=5)
_Decode/100000_samples-32    150.7Mi ± ∞ ¹   127.8Mi ± ∞ ¹  -15.18% (p=0.008 n=5)
_Decode/1000000_samples-32   1.422Gi ± ∞ ¹   1.185Gi ± ∞ ¹  -16.66% (p=0.008 n=5)
geomean                      13.93Mi         11.72Mi        -15.91%
¹ need >= 6 samples for confidence interval at level 0.95

                           │   /tmp/old   │              /tmp/new               │
                           │  allocs/op   │  allocs/op    vs base               │
_Decode/100_samples-32       2.839k ± ∞ ¹   2.239k ± ∞ ¹  -21.13% (p=0.008 n=5)
_Decode/1000_samples-32      28.05k ± ∞ ¹   22.05k ± ∞ ¹  -21.39% (p=0.008 n=5)
_Decode/10000_samples-32     280.1k ± ∞ ¹   220.1k ± ∞ ¹  -21.42% (p=0.008 n=5)
_Decode/100000_samples-32    2.800M ± ∞ ¹   2.200M ± ∞ ¹  -21.43% (p=0.008 n=5)
_Decode/1000000_samples-32   28.00M ± ∞ ¹   22.00M ± ∞ ¹  -21.43% (p=0.008 n=5)
geomean                      280.9k         220.9k        -21.36%
¹ need >= 6 samples for confidence interval at level 0.95

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]

[yeya24]

Thanks, looks good!

[yeya24]

I guess it could be beneficial to preallocate the buffer here, but we can leave it as it is now.

[alanprot]

yeah.. i just did not know the number of samples beforehand... but maybe put something like 100 should be ok?

[yeya24]

Yes a default value is fine.