[BUG] Non-reproducible search for restraint atoms

[lohedges]

Searching for restraint atoms using system.getRestraintAtoms("backbone") gives non-reproducible results due to the use of a set to hold the protein and nucleic acid residue names. This has been causing intermittent test failures in the CI for some time. Switching to a list for the container fixes things and the tests reproducibly pass.

One concern is that the search result appears to be order dependent. I'm not yet sure if this is a bug, or whether Sire has an upper limit on things resname searches where the items are comma separated. If so, then we can't guarantee the result. (Note that the examples in the tests do pass.)