BUG: Fix bug with reading dig strings

[larsoner]

Closes #13079

I expect the code to work, but I still need a file locally to test. Not sure if it's worth adding a tiny one to the testing dataset. I might try using tmp_path and coding one quickly instead. Draft until I can at least verify it works.

Also I noticed some code dup in read_fiducials so I removed it, and hit a bug when using mne show_fiff where getformat isn't a method of csr_array so we can just report the class name instead in show_fiff.