Add `base2.fixed.div` lowering with multiplicative inverse

[KFAFSP]

On closed Z groups, we can implement division by a constant divisor by computing a constant multiplicative inverse that we can multiply a quotient with. This optimization is necessary to achieve good fixed-point division performance, and is common in mainline compilers already.

• Are there parameters to this process? The precision of the inverse may be adjusted. How does the lowering derive these paramters? (Maybe introduce a base2.fixed.reciprocal operation for this).
• Implement a lowering from base2.fixed.div (maybe also base2.div) to multiplication for constant rhs terms.