Make the collector exporter mutate data, and remove unnecessary CopyTo

[dashpole]

Looking at #890 (comment)

This change makes the collector exporter now mutate data. We were doing multiple layers additional copying in our exporter to prevent mutating data, but it is likely better for us to just be a mutating exporter, and let the framework handle it.

This keeps the collector code clean, and should have better CPU/Memory usage (even after the additional copy) since we are copying fewer times.

Benchmarks (including those added in #893):

goos: linux
goarch: amd64
pkg: github.com/GoogleCloudPlatform/opentelemetry-operations-go/exporter/collector/internal/normalization
cpu: AMD EPYC 7B12
                                              │   main.txt    │             mutating.txt              │
                                              │    sec/op     │    sec/op     vs base                 │
NormalizeNumberDataPoint-2                      1845.0n ± ∞ ¹   532.6n ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeHistogramDataPoint-2                   2314.0n ± ∞ ¹   615.7n ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeExopnentialHistogramDataPoint-2        2071.0n ± ∞ ¹   673.7n ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeSummaryDataPoint-2                     1718.0n ± ∞ ¹   465.9n ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeNumberDataPoint-2                 2094.0n ± ∞ ¹   826.3n ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeHistogramDataPoint-2               2.565µ ± ∞ ¹   1.043µ ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeExponentialHistogramDataPoint-2   2148.0n ± ∞ ¹   869.4n ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeSummaryDataPoint-2                1775.0n ± ∞ ¹   836.2n ± ∞ ¹        ~ (p=1.000 n=1) ²
geomean                                          2.049µ         709.9n        -65.36%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                                              │   main.txt   │             mutating.txt             │
                                              │     B/op     │    B/op      vs base                 │
NormalizeNumberDataPoint-2                      408.00 ± ∞ ¹   16.00 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeHistogramDataPoint-2                   536.00 ± ∞ ¹   32.00 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeExopnentialHistogramDataPoint-2        600.00 ± ∞ ¹   48.00 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeSummaryDataPoint-2                     320.00 ± ∞ ¹   16.00 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeNumberDataPoint-2                  504.0 ± ∞ ¹   128.0 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeHistogramDataPoint-2               745.0 ± ∞ ¹   256.0 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeExponentialHistogramDataPoint-2    792.0 ± ∞ ¹   256.0 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeSummaryDataPoint-2                 432.0 ± ∞ ¹   128.0 ± ∞ ¹        ~ (p=1.000 n=1) ²
geomean                                          520.6         67.33        -87.07%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                                              │   main.txt   │             mutating.txt             │
                                              │  allocs/op   │  allocs/op   vs base                 │
NormalizeNumberDataPoint-2                      10.000 ± ∞ ¹   2.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeHistogramDataPoint-2                   13.000 ± ∞ ¹   3.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeExopnentialHistogramDataPoint-2        14.000 ± ∞ ¹   4.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
NormalizeSummaryDataPoint-2                      8.000 ± ∞ ¹   1.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeNumberDataPoint-2                 12.000 ± ∞ ¹   4.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeHistogramDataPoint-2              17.000 ± ∞ ¹   7.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeExponentialHistogramDataPoint-2   14.000 ± ∞ ¹   4.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
ResetNormalizeSummaryDataPoint-2                11.000 ± ∞ ¹   4.000 ± ∞ ¹        ~ (p=1.000 n=1) ²
geomean                                          12.09         3.191        -73.61%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

[codecov]

Codecov Report

Attention: Patch coverage is 54.44444% with 41 lines in your changes missing coverage. Please review.

Project coverage is 63.03%. Comparing base (4caace7) to head (c748e16).
Report is 43 commits behind head on main.

Files with missing lines	Patch %	Lines
...ctor/internal/normalization/standard_normalizer.go	61.11%	28 Missing ⚠️
...ctor/internal/normalization/disabled_normalizer.go	0.00%	12 Missing ⚠️
exporter/collector/metrics.go	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #892      +/-   ##
==========================================
+ Coverage   61.03%   63.03%   +1.99%     
==========================================
  Files          56       57       +1     
  Lines        5903     5967      +64     
==========================================
+ Hits         3603     3761     +158     
+ Misses       2143     2044      -99     
- Partials      157      162       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dashpole]

depends on #893

[aabmass]

It's weird the benchmarks now have ± ∞ ¹ vs #890 which had meaningful bounds. Any idea what's going on?

[dashpole]

I only did -count=1 because I was running short on time. I can re-run if you want

[aabmass]

Where is the actual flag indicating that the exporter is mutating? I'm guessing collector-contrib in a separate PR

[dashpole]

It is an option to exporterhelper: https://pkg.go.dev/go.opentelemetry.io/collector/exporter/exporterhelper#WithCapabilities, which we will need to set in our exporters