Use i8* instead of void* for generated LLVM IR

It looks like LLVM doesn't like using void* for opaque pointers. Trying to compile some LLVM IR actually gave me a useful hint:

```
$ /usr/local/Cellar/llvm/4.0.0_1/bin/opt add.ll
/usr/local/Cellar/llvm/4.0.0_1/bin/opt: add.ll:6:22: error: pointers to void are invalid - use i8* instead
define i32 @main(void*) {
```

Apparently, at least until [opaque pointer types](http://llvm.org/devmtg/2015-10/slides/Blaikie-OpaquePointerTypes.pdf) come to LLVM IR, the right type to use for this sort of thing is a pointer to 8-bit integers.

One easy way to do this would be to define a constant for the i8* type (called `FUNCTION_ENV_TYPE`?) and then use it wherever we currently have `llvm.PointerType.create(llvm.VoidType.create(), 0)`.

Search the repository for uses of llvm.PointerType.create(llvm.VoidType.create(), 0) and inspect the generated LLVM IR around those entry points. Replace the relevant void-pointer type uses with an i8* type, then run the generated IR through LLVM's opt command to confirm the pointers-to-void error is gone.

Written by the indexing model from the issue text.