Update JavaGCFlagChecker to recommend G1GC for Java 17+

[NickFirmani]

tl;dr g1gc is faster in Java17

[runningcode]

Thanks @NickFirmani do you have any benchmarks or information on in which scenarios G1GC is faster?

[runningcode]

Looks great. Thanks for adding the appropriate tests and updated the documentation

[runningcode]

On second thought, for migration purposes, let's leave the old flag around as well, see my comment.

[runningcode]

thanks for updating the verbiage here from "warn" to "fail" this is a breaking change but I agree with it.
can you also keep the old extension around (along with the new one) with a deprecated flag for backwards compatibility? we'll remove it in a future version but I want to make it easy to migrate.

[NickFirmani]

For sure, I added the old extension back, but set it to default to false. I think that makes sense, as a consumer of the library who updates will expect new build failures due to improvements they can make to their build.

[NickFirmani]

Bumping this up @runningcode, ty!

[runningcode]

This article seems to suggest that the parallel GC still gives the best performance: https://dev.to/cdsap/using-parellgc-for-the-kotlin-process-in-android-builds-2geh

[NickFirmani]

Yeah, I saw that article, which is super interesting, because for us, G1 was quite a bit faster, perhaps due to some quirk of our environment

[PaulWoitaschek]

What about modifying the logic that it checks that at least one is present?
In the end it's often project specific but having one of them is usually a good thing.

[sgammon]

@runningcode i believe that article points out that it is running pre-jvm17

Finally, as you know, with AGP 8, JDK is set to 17. I will publish a new article covering this experiment with the JDK 17 Garbage collector results.

I almost didn't catch it...