Replacement parser kernel

[dwightguth]

This PR essentially deletes the "parser" part of the parser kernel and replaces it with a new parser algorithm which is easier to understand and maintain. It also includes several smaller refactorings that arose during this process. Functionally, the old and new parser kernels behave identically. In terms of performance, on its own, the new parser kernel may be slower than the old one in a few cases, but in those cases which we have observed, there is usually a very massive nest of ambiguities, which we are able to postprocess faster due to the changes in the commit titled "move all priority filters into a single pass for performance". Overall the time spent parsing in large semantics is on par or faster with this branch on all the semantics I have tested.

I tried very hard to document the new algorithm so if anything is hard to follow, please let me know and I can add additional documentation. Bear in mind it's not likely to be completely understandable without reading the two papers I linked to in the javadoc of the EarleyParser class, though.

This has been tested against EVM, WASM, Michelson, and C in addition to the regression test suite.

[radumereuta]

This file can also be deleted: https://github.com/runtimeverification/k/blob/master/kernel/src/main/java/org/kframework/utils/algorithms/SCCTarjan.java

Overall pretty good. Indeed seems more straightforward. The previous implementation had other things in mind which didn't get implemented. The code seemed obfuscated because of that.