Add per page unicode mapping stats to the metadata in the PDFParser

As part of TIKA-2749, it would be useful to gather stats on characters that did not have a unicode mapping. Users could use this information now to determine which pages might benefit from OCR.

I propose two parallel arrays of ints, with one entry per page. The first would contain the count of # of characters per page, and the second would be a count of the unmapped unicode characters per page.

Many thanks to tilman for the guidance on how to gather this info easily.