Draft: Work on LL parser generator

Currently we do bottom up parsing using Bison. It's quite difficult to avoid conflicts in the grammar, see !736 (merged), where we eventually figured out how to do it, but it was a lot of work.

As an alternative, I am exploring a top down parser in this MR, which might be easier to develop. We would have to benchmark its speed against our current parser.

Given that the Bison parser seems to allow us to deliver on a minimal viable product, we will use it for now. But this MR is open to explore a long term solution.