Guido pointed out to me that some details about the per-interpreter state for the builtin types aren't especially clear. I'm addressing that by:
* adding a comment explaining that state
* adding some asserts to point out the relationship between each index and the interp/global runtime state