Cranelift: Simplify all our various CLIF load and store instructions

* `{u,s}load{8,16,32}` these are all equivalent to `load.i{8,16,32}` followed by `{u,s}extend`. We should just use that larger sequence and backends that have special lowerings for `uload8()` that are better than `uextend(load8())` should add rules to match the compound pattern.

* Ditto for `{u,s}store{8,16,32}`.

* Same for our SIMD loads like `uload16x4` et al. These can be `load.i16x4` followed by a lane-wise `uextend`, etc.