TODOs from hocr-check

[kba]

These are commented as FIXME at the end of hocr-check, I'll put them here for discussion.

• [ ] containment of paragraphs, columns, etc.
• [ ] ocr-recognized vs. actual tags
• [ ] warn about text outside ocr_ elements
• [ ] check title= attribute format
• [ ] check that only the right attributes are present on the right elements
• [ ] check for unrecognized ocr_ elements
• [ ] check for significant overlaps
• [ ] check that image files are not repeated

Keep this in check with hocr-spec (cross-reference maybe) and consider creating an XSD schema for use in ocr-fileformats (though these tend to be inflexible).

[zuphilip]

Copy the comment from #50 here:

See https://github.com/tmbdev/hocr-tools/pull/45#discussion_r77298171 . Also the tests should then be improved.