Add scalar size specifiers

[daboehme]

This adds scalar size specifiers like i16, u8, f32, etc. to the type system. Addresses #30.

Primarily this allows us to use smaller-size scalars in compound types in the C API. This wasn't possible before. For example, we can now import a list of floats like so:

float floats[4] = { 1.5, 2.5, 3.5, 4.5 };
adiak_namevalue("floats", adiak_general, NULL, "{%f32}", floats, 4);

Internally any shorter types are converted to double or int, so e.g. inputs passed as i16 are still stored as adiak_int. This is consistent with the C++ API and the way scalars are handled as varargs in C. It also requires no changes on the tool side. However since we know the real size of the input type we can address them correctly through void* pointers.