- improve tesserocr-batch: add multiprocessing
  and probability / confmat output
- new script dta-txt2gt for DTA text-only GT
  (including line wrapping with hyphenation)
- new script ocrd-gt2pkl to reduce METS/PAGE
  workspaces into plaintext pickle dumps
  (including alignment)
- new script prob2pkl to combine OCR results
  for plaintext string and probabilities into
  pickle dump format
- new script confmat2pkl to combine OCR results
  for plaintet string with alternatives and
  probabilities into pickle dump format
- add proper module installation

- faster (batch processing, parallel)
- tesseract OCR uses custom CLI (for batch processing, for confidence output)
- added ocropus OCR (2 models)
- generate confidence output (1-best only), encapsulate as tuples into pickle files