add INLINABLE directive

[stylewarning]

This tells Coalton to save the AST for future inlining. It might be a no-op now but it will be valuable as we try to scale Coalton up to larger code bases.

(inlinable)
(define (f x) ...)

I'm also open to other syntax, like

(inline :always)    ; current INLINR
(inline :maybe)     ; no equivalent, let the compiler decide if it thinks inlining is good
(inline :able)      ; this proposal: INLINABLE

Or something like that.

[garlic0x1]

Since :maybe is essentially the default due to the heuristic inliner, maybe we can just add (inlinable) as well as (noinline) to keep the current syntax so the common case of just declaring inline stays short

[garlic0x1]

or do you mean that :maybe would use heuristics to emit declaims for lisp callers?

[stylewarning]

When the heuristic inliner is on, it has the effect of making every single function

(inline :maybe) 

but again in the long term assuming every function can always be inlined is not scalable. (Imagine compiling 5 million lines of code and every single definition is getting saved and analyzed.) It's nice to have an option to signal to the compiler that you think something is worth inlining, but to use the context of the individual calls to make that decision.