Follow up from #3132 (https://github.com/MDAnalysis/mdanalysis/pull/3132/files#r632672743)
The way we use unitcell in the PQRReader isn't super consistent given that we really only set it to None. Given the way the format works (https://github.com/MDAnalysis/mdanalysis/pull/3132/files#r629707068), I think we can just safely remove any code paths that assume non-empty unitcell values.