testkit: extract targetroot from scalacOptions

[bjaglin]

Second take at #1275 leveraging #1389 (with the addition of the first preliminary commit, allowing the targetroot to be repeated in the CLI when dealing with a complex classpath)

Unlocks scalacenter/sbt-scalafix#219

[bjaglin]

unreleased

[mlachkar]

I didn't think about having multiple targetroot directories.
I do prefer this approach for specifying semanticdbTargetRoot.

[bjaglin]

I didn't think about having multiple targetroot directories.

Me neither at first, and I didn't expect this PR to take me there (since it's about inferring it from scalacOptions in testkit, meaning that it can't be repeated) but I think it demonstrates how/why having a separate CLI argument makes sense: specifically when the input is so complex that it was compiled in several scalac invokations. Of course, in that case we would like to provide several "sets" of scalacOptions, but that's another story.

We might want to expose semanticdbTargetroots as a property in testkit to handle such use-cases, but I thought we could live without it for now and revisit later. I think it's fair for the testkit to follow what the CLI has been doing until now anyway.

[mlachkar]

Ok, let's do that like this, but I think if we start using withTargetroots in the sbt-plugin, we should do the same in the testkit, even if we retrieve the targetroot from the scalacOptions for now.