Some suggested follow-ups from gh-3935:

1. Add support and testing for pathlib object handling to SingleFrameReaderBase and/or upstream this to the proto infrastructure. The pathlib handling/testing in the connected PR is restricted to ReaderBase for now.
2. The testing added in the above PR might eventually benefit from being consolidated to i.e., universe creation over a parametrized set of trajectory formats instead of the copy-paste sampling I did for GRO and DCD formats.