Tests: Better pretty print

[RunDevelopment]

This is a major improvement for the pretty-printer.

The main change is that TokenStreamTransformer.prettyprint now takes a Prism token stream instead of a simplified token stream. This means that it has a lot more information to work with and it uses that information.

It mostly behaves like the old pretty-printer with 3 main differences:

1. Tokens that contain a nested token stream that only contains a single string are printed as one line.
   Example: ["namespace", ["java"]]
2. The pretty-printer can now glue tokens together so that they are printed in a single line.
   Example: ["number", "100"], ["operator", ".."], ["number", "100"]
3. Line breaks in the input code may now cause empty lines in the output. Previously, only blank strings containing more at least 2 line breaks produce empty lines. This helps readability as it groups tokens together.

If anyone wants to know about how it works in detail, I'll explain, but it's a bit much so I won't do so here.

---

I also want to ask whether I should reformat all .test files.
The pretty-printer is quite good and it would be nice if all test files were formatted consistently. However, 50% of all test files change when reformatted.