Boundary trial moves

[joaander]

Add trial moves that change the simulation boundary.

As with LocalTrial and Sweep, boundary moves will likely benefit from a design where individual trials (scale, shear, aspect, etc...) are implemented separately from the overall code that manages the preparation and application of the moves. Review the HOOMD-blue code for BoxMC to see how the acceptance criteria vary slightly for the different types of moves.

We will need to add some new traits to the boundary types to facilitate the different types of deformations. These traits will both need to deform the boundary and describe how bodies in the boundary are deformed. Not all boundaries will support all types of deformations (e.g. Cuboid cannot be sheared), so they will need to be broken up appropriately.

[janbridley]

I’ll spend some time on this once I’m back!

[joaander]

Before boundary moves can be implemented, we need to solve an API design issue around setting boundaries. Details here:

hoomd-rs/hoomd-microstate/src/microstate.rs

Lines 503 to 512 in fd6ba05

	TODO: Replace with setter. `boundary_mut` allows the caller to create an
	invalid microstate by changing the boundary in such a way that sites may
	be outside. Changing the boundary will also require regenerating ghost
	sites. Just checking for a valid boundary on set will pose some difficulty
	to the caller. To increase the boundary, the caller will need to set
	the new boundary and then move the bodies. To decrease the boundary, the
	caller will need to move the bodies and then set the boundary. Perhaps a
	`set_boundary_and_update_bodies` method that does both simultaneously would
	solve this? It could take a function that updates the bodies along with the
	new boundary.

The biggest part of this decisions is whether to allow boundaries to be mutable or are they transformed with functions that generate new boundaries? In either case, mutating or creating new Periodic boundaries is fallible. The boundary may be made so small that the maximum interaction range becomes greater than the maximum allowable interaction range.

[janbridley]

Set boundary for MC: if this needs to create a whole new Microstate, we don't need to change the boundary. We just create a whole new one, evaluate acceptance, and then assign or discard the microstate back to the original variable. Instead of taking in a mutable Microstate and mutating it, we take in an owned and return as above