Processor.input_files: handle list of pageId

[kba]

No description provided.

[bertsky]

Sorry, I am just seeing this now: We actually have to rewrite the whole function. Up above, we filter by MIME type alone. So whenever (for some list of pageIds) there is even a single page with PAGE annotation, only that page survives.

What we need instead:

• find_all_files filtering MIME for PAGE or image/*
• aggregate results by page ID, and iterate:
• if the page has a PAGE hit, keep only that
• else if it has a single hit, keep it
• else raise

[bertsky]

Superseded by #635