Vary snapshot output dir based on applied kotlin plugin

[jrodbx]

Closes #595

[jrodbx]

Closing, to think about this a little more.

Moving away from withType callbacks results in spaghetti code around determining snapshotSourceSet depending on which plugin is applied. Using hasPlugin synchronous code causes plugin ordering issues.

I think the plugin code might need some more refactoring to better enable this.

[jrodbx]

question for @JakeWharton @TWiStErRob: should we require a kotlin plugin to be added? I think right now, we haven't required it, but it might be antiquated to support this?

[JakeWharton]

Not a huge fan of forcing it when it's not strictly required. There's absolutely people still writing Java that will use Paparazzi.

[TWiStErRob]

Agreed. Support it, but don't require it.
What would be the reason to require it?

I think requiring near-latest AGP is already a higher than necessary barrier to entry, but it's a good tradeoff for maintenance and moving fast.

[autonomousapps]

I'm still strongly in favor of using compileOnly for dependencies on other plugins. I note that this plugin has an implementation dependency in KGP; is that necessary? That'll put KGP on the build classpath even if the project using Paparazzi doesn't use Kotlin.

The other issue with having strong dependencies on other plugins is the way Gradle's classloader hierarchy works. Depending on where this plugin gets added to a build, it can complicate things for users who want to use different versions of the other plugins.