[stdlib] Fix allow implicit casting MutableOrigin <-> ImmutableOrigin

[martinvuyk]

Fix allow implicit casting MutableOrigin <-> ImmutableOrigin by adding implicit constructors to and fro for Span and StringSlice

[nmsmith]

Small note: If we end up modelling origin mutation as an effect that functions perform, rather than today's approach where structs and functions are parameterized by is_mutable parameters, this PR would end up being redundant, because "mutable slices" and "immutable slices" (etc.) would have the same type. (This idea has been in discussion for a while.)

That's not to say this PR shouldn't be accepted. Maybe this is solving some near-term problems that people have been experiencing. I'm out of the loop on this.

[martinvuyk]

Hi @nmsmith, I would love for this PR to become redundant. Parametric origins are a pain. This is mostly to facilitate work in function signatures and at call-site to allow using both types interchangeably on both ends and letting implicit construction do the job of casting