refactor: improve the tck process to be lazy evaluated

[jdaugherty]

I was speaking to https://github.com/vampire in the Gradle slack and he had some feedback on my original tck implementation. I think I implemented them the way he suggested so they're lazy and the task does not occur at configuration time.

Some highlights of that conversation:

1. configurations are just file collections, so you can take advantage of this.
2. If only one dependency is exported, then singleFile is a way to avoid looping over files / hard coding a name check.
3. The original code was doing configuration resolution at configuration time which has known drawbacks.
4. It's better to use an artifact transform or instead of using tasks like Copy/Sync, use copy { ... } or sync { ... } in a doFirst or doLast block.
5. resolve() is 99.87% of the time the wrong solution.
6. Copy or copy { ... } should almost 98.7% of the time be Sync or sync { ... } to not risk leaving stale files lying around.

@matrei with this new found education from an amazing developer, I thought I'd improve my original implementation =)

[matrei]

Awesome!

[jdaugherty]

I need to make 2 more changes, but can't until later today:

Revert the changes to the tck gradle file and change the configuration to transitive=false.

Remove the dependsOn and add an input to the configuration instead

[jdaugherty]

@matrei These changes are done, if the tests pass and Vampire comes back to me saying it's ok, I intend to merge. He also reminded me that dependsOn() is a smell and that it's better to set the inputs of a task. Take a look at my latest iteration.

[jdaugherty]

@matrei as discussed, I'm going to go ahead and merge this. Let me know if you have any other feedback!