I think lowering vars to db indices works now

hir2 lowering and var conversion is slowly coming together

and even might work properly!

ok I have some meaningful tests for de bruijn index lowering now

Have I finally come up with a construction that doesn't make me cry?

Ok I'm still procrastinating about de bruijn indices.

Not sure what to do about this.

grmbl grmbl symbol tables

ok I still think this will work

Puttin' more work into avoiding standard de bruijn index construction

Mostly to prove I can

Hokay starting to remember wtf I was doing

Ok we now give lowered variables indices

We MIGHT even do it correctly!

resolving names into de bruijn indices is the worst kind of problem:

uninteresting, meticulous and difficult.  Welp, I tried.

Ok I just ripped a bunch of type ambiguity out of hir2

and *hnnnrg* does it feel good.  So it should now be *very clear*
where type abstraction and type application occurs, and where type
variables are declared.  And nothing new is broken, as far as I
can tell!

Next up I think it's time to stuff a symbol table into the HIR2
lowering process, properly translate vars to indices, and then we
can start evaluating shit.

oh actually compiling first is a good idea

Slowly refactoring kinds into actual arg lists

Continuing along with eval_type()

Still not sure where to fit unification into this, it's a little
sticky.  I think eval_type() creates my unification variables and
then quote() or type_equiv() or type_of() just resolves them.

I think it's time to RIP AND TEAR my existing type infra though, which will
hella sting.  Maybe I can just duplicate it in parallel still?

Ok I think I've broken it down to a test case to start on unification

Ok it's time to think Hard about unification vs bidi

I think I'm going to try out unification.  reconstruct()
is basically quote() and having eval_type() be able to introduce
unknowns which normalize() or something then tries to eliminate
Feels Right.  I already have a type_equiv() function that is
essentially used wherever the old type checker would use unify()...

Gnawing on the boring parts of the NbE typechecker

Refactoring out HIR2 lowering and thinking about names/indices.

already have an interesting problem ig