bug: pan fails saving on one-character author field

Symptom:
Pan attempts to save a group's headers and fails with log message "Unable to save \"<group>\" Success". No debug output is generated. The .pan2/groups/<group>.tmp file remains.

Once this occurs Pan is stuck. New headers are visible only in the session they are downloaded, and will not be downloaded again by Get New Headers.

What goes wrong:
Pan maps characters [a-zA-Z0-9!@#$%^&*()] to frequent authors to save disk space when saving out headers. When loading a header from disk with a single-character author field, Pan assumes this is one of the mapped frequent authors. This is almost certainly wrong if the author field really is a single character.

The bug manifests when loading headers from file where a legitimate single-character author field is not mapped to a frequent author in the file. One example seen in the wild has author field "-", not one of the mapped characters. Pan correctly saves this in the header file on the initial download, but after loading headers from the file can't resave them.

Daisy's lazy fix:
If a single-character author field doesn't resolve to a frequent author then use the character as the author. The character-to-author scheme is fatally flawed if a character is a valid author anyway.

There's more:
I have about two dozen fixes in my local branch for reported and unreported bugs. Mostly trivial stuff like this. Just ask. I'm afraid this project is not really maintained any more.

Queue Duncan mansplaining. With footnotes.