Merge Int64DataPoint and DoubleDataPoint into ScalarDataPoint and support different sum types in Histogram

[bogdandrutu]

This is the first option to fix #206

Cannot see any significant difference between merged messages vs split messages, code is here https://github.com/bogdandrutu/metrics-proto

goos: darwin
goarch: amd64
pkg: github.com/bogdandrutu/metrics-proto/benchmark
BenchmarkEncodeDecode
BenchmarkEncodeDecode/merged
BenchmarkEncodeDecode/merged-16         	   34406	     33744 ns/op	   30288 B/op	     523 allocs/op
BenchmarkEncodeDecode/unmerged
BenchmarkEncodeDecode/unmerged-16       	   36130	     33190 ns/op	   29168 B/op	     523 allocs/op

[jmacd]

I prefer #209 as it uses less memory, but I also support this option as it tends to produce smaller code.

[james-bebbington]

I would value the readability & maintainability of this proto going forward (which might allow faster iteration & increase adoption) more highly than a <1% performance difference in this case, so prefer this over #209. I find easier to read this than the other one, and I would imagine the other approach would get messier (more duplication) as we add more to it.

(but I don't feel that strongly about it either way)

[jmacd]

@jkwatson put in a vote for this PR in Gitter. I'm on the fence, but I should mention that I proposed a similar PR to this one in the past: #172

I would prefer that this is settled than to debate these options, though. Can we move forward with this version?

[tigrannajaryan]

I agree performance difference is negligible and we need to choose based on usability or some other criteria.
@bogdandrutu can you also assign the value to measurement_value_type where appropriate to make the code complete?

[bogdandrutu]

Thought about this more and I prefer #209 for the reasons I put in gitter: extensibility (can add new types that work only on some measurement types like strings if we decide to support that); no corner cases like measurement type is long and exemplars are using double values.

It seems there is no clear winner so I would want to move forward with that one

[bogdandrutu]

I am going to close this and ask everyone to review #209.