[processor/transform] Add 'transform.flatten.logs' feature gate

[djaglowski]

This PR proposes a feature gate which would enable the Flatten/Unflatten behavior described here.
This was discussed in the Collector SIG recently and it was mentioned that a feature gate would be a reasonable way to implement this.

One immediate question: Should this be purely a feature gate, or should there be a config option on the processor which fails Validate if the feature gate is not set?

[BinaryFissionGames]

I'd advocate for adding in a config option in addition to the feature flag; I think it can help with slowly migrating configs over to this option if needed for more complex configurations. In addition, I think in some scenarios where you don't need the flattened representation, it would be nice to avoid the overhead of having to flatten/unflatten.

It also makes it easier to swap between the hierarchy and flattened representation for agents managed by OpAMP (feature flags are currently treated as static and not configurable through OpAMP, at least in the current supervisor design).

[crobert-1]

Should we add a note here as well that the config option is also required to enable this? It may be clear enough from the README, so just a suggestion.

[evan-bradley]

I think a note about the config option would also be good in the subtext, just so users hopefully have enough info to use the feature from the changelog entry.

[TylerHelmuth]

@djaglowski Can you add a benchmark test for this feature? I know we need it in order to produce the right outcomes for certain scenarios, but I want to be able to claim it is performant when used.

[djaglowski]

I added a simple pair of benchmarks, one with and the other without the flatten option. It's definitely a performance hit but seems reasonable to me for when correctness is needed. Results on my machine:

BenchmarkLogsWithoutFlatten-10    	  389914	      2616 ns/op	     241 B/op	       8 allocs/op
BenchmarkLogsWithFlatten-10       	   63205	     19262 ns/op	    2726 B/op	      53 allocs/op

[crobert-1]

Thanks for adding the tests!

[evan-bradley]

Looks good to me. I think the feature gate is sufficient to get this in despite the early state of the functionality.

[evan-bradley]

I think a note about the config option would also be good in the subtext, just so users hopefully have enough info to use the feature from the changelog entry.