Reproducible tests on different processor deployments

### Are there any linked Issues or Pull Requests?

_No response_

### Brief description

It is plausible to create tests of some science and technical capabilities that result in bit-identical results when run on different processor deployments (MPI ranks or shared-memory threads). These would be supported by:

* reproducible global reductions
* constraining the order of calculations of shared dofs (in continuous fields).
* suite capability to configure such tests

Possible low-hanging fruit would be tests of the lfric_apps physics codes which, largely, do not use continuous fields.

### Further details of the issue.

_No response_

Start by reviewing the lfric_apps physics codes and the test suite configuration. Investigate how tests could use reproducible global reductions and constrained shared-DOF calculation order; done means selected tests produce bit-identical results across MPI ranks or shared-memory threads.

Written by the indexing model from the issue text.