Cleanup frontends for LM, C, and LSTS through until type checking

[andrew-johnson-4]

In preparation for a "reasonable" standard of functionality.

TODO:

• ✓ port LSTS tokenizer to LSTS
• ✓ port C tokenizer to LSTS
• ~port types to use smart strings~
• port unify / can-unify to LSTS
• ~port AST to use SmartStrings~
• port LSTS frontend to LSTS
• port C frontend to LSTS
• port LM frontend to LSTS
• port preprocessor to LSTS
• port typechecker to LSTS
  • add support for distinct arity per type Tuple<x,y> vs Tuple<x,y,z>
  • add support for multipass inference with Algorithm W
  • 100% error message coverage with good errors
  • add support for corollary functions

everything through to typechecking should use LSTS at this point (and be good idiomatic code)