Generalize Literal::externalize/internalize for strings and shareability

[kripken]

Is it worth testing shareability here? (is the fuzzer ready?)

[tlively]

It is worth testing shared i31s (which will reveal that the assertion at the beginning of Literal::externalize() needs to be updated), but the fuzzer does not support shared stringref yet.

[kripken]

Ok, I added testing for shared i31. That found a few more places that needed shareability fixes.

[tlively]

One thing to note is that now that we are closer to being able to run and directly import the upstream spec tests, any additional testing we add to our own spec tests will be able to eventually benefit everyone.

[kripken]

ASan tests found another issue with shareability, in a destructor. Fixed.

@tlively I am getting a little worried that we may have many more places with x.getHeapType() == HeapType::y that ignore shareability. Perhaps we should audit them all somehow?

[tlively]

Good idea, I will take a look.