Rework word-count exercise to be consistent #445

In 2015 we decided to make the word-count exercise more interesting and not only split on white spaces (#177) which was inspired by the Haskell track. Since than other tracks where against it and so Haskell moved back to white space only (exercism/haskell#501). This is also the way the canonical test data looks like. So maybe we should drop the python specialty to be more uniform.