prometheusremotewriteexporter: Remove unnecessary buffer copy in proto conversion

[bogdandrutu]

This happens because the proto message is generated with gogoproto, and it will go on the path https://github.com/gogo/protobuf/blob/f67b8970b736e53dbd7d0a27146c8f1ac52f74e5/proto/table_marshal.go#L2977

Benchmark numbers show improvement as well.

After:

BenchmarkExecute/numSample=100
BenchmarkExecute/numSample=100-12         	   21752	     52500 ns/op	    6635 B/op	      83 allocs/op
BenchmarkExecute/numSample=1000
BenchmarkExecute/numSample=1000-12        	   10000	    131918 ns/op	   23096 B/op	      86 allocs/op
BenchmarkExecute/numSample=10000
BenchmarkExecute/numSample=10000-12       	    1479	    886487 ns/op	   41477 B/op	      86 allocs/op

Before:

BenchmarkExecute/numSample=100
BenchmarkExecute/numSample=100-12         	   20642	     55670 ns/op	    6635 B/op	      83 allocs/op
BenchmarkExecute/numSample=1000
BenchmarkExecute/numSample=1000-12        	    8640	    145875 ns/op	   23121 B/op	      86 allocs/op
BenchmarkExecute/numSample=10000
BenchmarkExecute/numSample=10000-12       	    1015	   1257688 ns/op	  304366 B/op	      87 allocs/op

[jmichalek132]

LGTM.

[mx-psi]

This seems like the most likely cause of #42414. Should we revert until we can fix that?

[jmichalek132]

This seems like the most likely cause of #42414. Should we revert until we can fix that?

I dropped your question into slack to get feedback from the maintainers of the prom components.

[jmichalek132]

This seems like the most likely cause of #42414. Should we revert until we can fix that?

I think that's a good idea but, I am not an approver for the component so maybe you might want to wait for a reply from them?

cc @ywwg @krajorama @ArthurSens @dashpole

[ArthurSens]

I'm sorry, but I couldn't find the time to review this. If you feel like reverting this one would solve the flaky tests, please go for it. If we notice that the flakiness is still ongoing after the revert, we can also revert the revert.

In an ideal world, merging PRs without approval from codeowners shouldn't happen. However, I feel like we have more PRs being opened than our current group of codeowners can handle, so I don't want to block PRs from collector maintainers because of our slow responsiveness.

I'm open to ideas here

[jmichalek132]

fyi raised the PR to revert it #42501