StashRepository: Merge init() into the constructor

[proski]

*  StashRepository: Merge init() into the constructor
   
   Jenkins documentation doesn't guarantee that start() would not be called
   after run() for a trigger.
   
   While it doesn't happen in freestyle projects, Jenkins would actually
   call start() after run() on the trigger in a pipeline project.
   
   That would create a new StashRepository object that would not have the
   StashApiClient object initialized. When the build completes, it would try
   to post the build result as a PR comment. That would throw an exception.
   
   To prevent it, remove the init() method. Move StashApiClient
   initialization inside the StashRepository constructor, so that the
   StashRepository object always has a valid StashApiClient object.
   
   Add WireMock as a test dependency. Check that the newly constructed
   StashRepository object would actually send an HTTP request.

That's one of the commits that make pipeline support possible. Extracted from #69 to help with review.

I realize that constructors are less testable than other methods, but the focus here is on fixing an issue with pipelines with minimal changes.

In the future, we can use factory pattern or generics to improve testability. Also, it might be possible not to replace the StashRepository object on the start() call for the same job. I'm not doing any optimization at this time, I'm making the code more robust.

[proski]

The requested changes have been made. The unit test is using WireMock now.

[proski]

Reworked the PR to keep the constructors short. As a side effect, the diff should be easier to read.

[proski]

Improved the comment for the constructor used in unit tests.

[proski]

Using Java 8 version of WireMock and dynamic HTTP port for testing.

[proski]

Added a verification that WireMock received a GET request.