Add in parse tree testing for every grammar, every input file

[kaby76]

This repo currently tests the outputted CST of a parse for a small subset of grammars. With new versions of Antlr coming to fore (different targets, e.g., Wasm, Rust; rewrites, e.g., Antlr4ng), it's important to make sure these produce identical parse trees, checking the correctness of the generated parsers, runtime, and ports. This PR adds in parse tree testing for all test input, for all grammars.

This fixes two issues:

• Format all .tree files #748
• We should always test the parse trees generated #2991

All .tree files are removed and replaced with .ipt "Indented Parse Tree" files. These files are human readable, with each node in the parse tree on a separate line and indented proportional to the depth of the node in the CST. The old Maven tester still works, but only for quick sanity checking of grammar changes. The old tester is no longer used for comparing the outputted CSTs. All input files that are parsed now have a corresponding .ipt file, which must be kept in sync with the grammar and test file.

The code for formatted parse trees is implemented in the driver code for testing in this repo. Antlr does not have any implementation of formatted parse trees, and likely never will, as it is difficult to get buy-in on any new features.

Going forward, people will need to update the .ipt with a change in the grammar. They can do this by:

• running trgen locally;
• or, running a build and copying and pasting the resulting trees to the .ipt's.

People will need be careful of the newline characters of both input and .ipt files generated by the parser. The .ipt files must be in Unix-style endings for easy "git diff" comparisons. Any input files for testing should probably be in "eol=lf" representation in the git repo. You will probably need to update the .gitattributes file.