[awsemfexporter] Restructure Metric Translator Logic

[kohrapha]

Description:
This PR restructures the code for building CW Metrics from pdata.Metric to remove duplicated logic. To do this, we had to define the following interfaces:

1. DataPoints that wraps pdata.(Int|Double|DoubleHistogram)DataPointSlice
2. DataPoint that wraps pdata.(Int|Double|DoubleHistogram)DataPoint

However, since the At(i) method for DataPoints requires returning a DataPoint, which is not the same as what pdata.(Int|Double|DoubleHistogram)DataPoint returns, we have to define a light conversion struct that provides the correct method signature.

The following functions have also been renamed to communicate its intended output better:

• getMeasurements -> getCWMetrics
• buildCWMetricFromDP and buildCWMetricFromHistogram are combined into buildCWMetric

This code change should not have any visible API changes. It's merely a reorganization of the code to clean it up and remove code duplication.

Testing:
This CR was run against the unit tests successfully.

Performance
I ran some benchmarks to compare the performance of this new code change against the old code, and got the following results:

benchmark                                                old ns/op     new ns/op     delta
BenchmarkTranslateOtToCWMetricWithInstrLibrary-12        9588          9182          -4.23%
BenchmarkTranslateOtToCWMetricWithoutInstrLibrary-12     8836          8475          -4.09%
BenchmarkTranslateOtToCWMetricWithNamespace-12           17.4          16.5          -5.17%
BenchmarkTranslateCWMetricToEMF-12                       23867         22283         -6.64%

benchmark                                                old allocs     new allocs     delta
BenchmarkTranslateOtToCWMetricWithInstrLibrary-12        131            122            -6.87%
BenchmarkTranslateOtToCWMetricWithoutInstrLibrary-12     116            112            -3.45%
BenchmarkTranslateOtToCWMetricWithNamespace-12           0              0              +0.00%
BenchmarkTranslateCWMetricToEMF-12                       45             45             +0.00%

benchmark                                                old bytes     new bytes     delta
BenchmarkTranslateOtToCWMetricWithInstrLibrary-12        7462          7190          -3.65%
BenchmarkTranslateOtToCWMetricWithoutInstrLibrary-12     6853          6869          +0.23%
BenchmarkTranslateOtToCWMetricWithNamespace-12           0             0             +0.00%
BenchmarkTranslateCWMetricToEMF-12                       84394         84391         -0.00%

We note that, not only is there no negative impact on the performance, we see a minor improvement across the board. This is great news because we managed not to impact performance while improving code readability for developers.

[linux-foundation-easycla]

The committers are authorized under a signed CLA.

• ✅ Raphael Koh (a587a19)

[kohrapha]

Tagging @mxiamxia, @hdj630, and @shaochengwang for visibility.

[anuraaga]

Just small nit but looks nice, thanks

[anuraaga]

Not sure but FWIU, it's still idiomatic to stick with append even when preallocating the array in Go.

[kohrapha]

Oh interesting, I can change it if it makes more sense. I just did some research during my implementation phase and got the sense that the general opinion on SO was to use indexing if you know the length ahead of time.

[bogdandrutu]

If you do this, you avoid checking for capacity all the time, so it is indeed a small improvement.

[anuraaga]

Seems like a great improvement - separate from this PR would be interesting if this is useful to other exporters and could be in collector core somewhere

[kohrapha]

Yea, would be useful to see if we can move this interface to collector core for exporters to use.

[bogdandrutu]

If you do this, you avoid checking for capacity all the time, so it is indeed a small improvement.