Add ability to specify non-uniform Pauli distributions for twirling

[ihincks]

Description

Currently, Pauli-twirling is hard-coded to use samplomatic.distributions.UniformPauli. In order to be able to implement strategies such as those described here, it would be nice to somehow be able to use a non-uniform sampling strategy. This particular paper discovers the distribution to use as a heuristic minimization function of an expected noise model, and IMO it would be out-of-scope for samplomatic to try and implement that particular minimization by porting the Mathematica scripts. However, we could provide tools at a more general level that enable users to implement these techniques.

Acceptance criteria

• Users are able to specify non-uniform Pauli sampling strategies on a per-box level.