Add multiply-high operation

[CatsAreFluffy]

Is your feature request related to a problem? Please describe.
Currently, multiplying numbers with more than 24 bits is awkward, since you can only access the lower 24 bits of the full 48-bit result.

Describe the solution you'd like
A new primitive operation that multiplies two numbers and returns the high half of the product.

Describe alternatives you've considered
Waiting for native support for bigger numbers (and then asking for multiply-high for those since people will always need bigger numbers)

Additional context

[enricozb]

We probably won't add this, and will instead just suggest using the extended numerics (#319) once support for that is added.

[VictorTaelin]

Wait why not? Looks like a great idea actually. Now that I think about it, a set of opcodes to allow for u48 implementations using (u24,u24) tuples look like an amazing intermediate between fast unboxed numbers and slow boxed ones. Perhaps we could actually implement native u48's as (CON u24 u24)?

[enricozb]

We could add it, but we'd be supporting quite a few numeric representations though.