adding conservative constant folding

[nevakrien]

this is currently just a rough sketch however it does already work for very simple cases (like minus_2.b)
I believe this specific way of doing it is guaranteed to be 100% correct for all b code everywhere although i might have messed up some of the actual computations

[nevakrien]

this attempts to solve #154

[nevakrien]

i went ahead and also added some dead code elimination to solve #155
it covers the case in the test code and again 100% conservative I am even letting users do some cursed things with asm if they want (labels within user asm are considered potentially jumpble and thus not dead)

[rexim]

This is a rather unsatisfactory state of affairs. I'd expect a bit more from a PR that introduces constant folding. I'd expect at least the level I demonstrated on the stream where I prototyped it https://www.youtube.com/watch?v=L-IzN9-fqWg

[nevakrien]

Yes that kinda requires playing with refrences to auto at least a little bit.

The dead code elimination is actually pretty good but for the constant folding I do need to start tracking the autovars which is more of a pain.

[rexim]

Yes that kinda requires playing with refrences to auto at least a little bit.

The dead code elimination is actually pretty good but for the constant folding I do need to start tracking the autovars which is more of a pain.

Understandable