Better support moving geometry

Some problems that currently crop up:

- [x] Recreating meshes is awkward. @alexfikl proposes `Discretization.copy(mesh=...)` (or `.replace(mesh=...)`). (which?) That seems easy and useful, but it requires remembering the group factory (probably OK).
- `Mesh`es can't be created efficiently. Their data structure involves `numpy` at the moment. Maybe meshes should start using array contexts, too?
- [ ] Group factories need to recompute discretization node arrays. If the mesh already uses the right unit nodes, this can probably be avoided. (#135 )

cc @alexfikl

Read the Discretization and Mesh implementations, then inspect how group factories build discretization node arrays. Review the related work in issue #135 and resolve which mesh-replacement API and array-context approach are intended. Done means the listed mesh creation, replacement, and node-recomputation problems have a documented, tested resolution.

Written by the indexing model from the issue text.