Feature/dram ag distill with new rule parser for distill, traits, and product

[madeline-scyphers]

feat: add new DRAM rule parser submodule for traits and distill
New python Lark base rule parser that defines rule grammer
for traits grammer that can be reused for distill and product.
This rule parser is more accurate and less error prone than
the older traits parser completely custom coded. THis fixes
a number of bugs where things were being double counted and not
parsed correctly. This also allows distill, traits, and product
eventually to all use the same rule parsing code.

Rule parser is implemented with polars instead of Pandas for a
few reasons. It is a bit faster, and it allows lazy query planning.
The annotation df is not currently lazy (eager rn), but with plans
to allow it to be lazy. Lazy DataFrames can be more memory efficient
by only loading the data/columns needed, and doing query optimization
to speed up and require less memory for intermediate steps.