bashlib: better tooling required to handle PAGE/images in same fileGrp

[kba]

c.f. https://github.com/OCR-D/ocrd_olena/pull/68/files this should be easier.

[bertsky]

…and core can also be more efficient!

This is probably not much more than a different mode for the find command: add the option --by-page to yield at most one file per pageId, preferably mimetype PAGE, or "image/*", but failing if there are more than 1 PAGE or image files. (Failure could mean simply skip that pageId, or exiting.)

IMO the best option would be to have a shared, fast implementation for both this CLI option and Processor.input_files.

[bertsky]

Or expose Processor.input_files directly to bashlib.

[kba]

Since we now have ocrd bashlib input-files is this solved?

[bertsky]

Yes (just forgot to link).