Getting two children of a node to match the same type isn't quite working

The global.set node now produces a 'get' after the 'set'. This leads to a more concise and traceable effect on the stack

Globals seem to have multiple types throughout generation

These node types are necessary for the type system, since they always result in an i32 regardless of the type of the operands

Adds two convenience methods: get-base-type-name and combine symbols

get-base-type-name attempts to fully concretize a type variable before getting its name

combine-symbols takes in two symbols and glues them together. This is useful for making instructions like 'i32.add', where the 'i32' could be any type

Move the complexity from the fresh property into the printer and only ask xsmith about choices that matter.

Section 6.4.6 in the wasm spec

IE Not just i32s.  I imagine you might require declarations to be
initialized with a literal value.  If that's the case, maybe you
should make a `LiteralExpression` node as a supertype of the various
literals.

The `case` form is terrible, and I wish it didn't exist.  I never use
it, and I've only seen new Racketeers use it wrong.  You never want
`case`, you want `cond` or `match` instead.

The `case` form only allows literal data in the match, so if you try
to use a variable it will be interpreted as a symbol instead!
Moreover, `case` and `cond` both have a terrible design mistake that
the fall-through case returns `void` instead of raising an error!
I've only seen `case` result in tiresome debugging.