[WIP] Add OCR support

[vinayak-mehta]

Closes #14

Output on this image-based PDF (much better than vanilla tesseract):

        0       1         2               3                4                5               6
0    u@ta          nictance                       Percent Fu       Savings el                
1    Name  (1lkm)      (mi)  Improved Speed  Decreased Accel  Eliminate Stops  Decreased Idle
2  2012 2    3.30       1.3            5.9%             9.5%            29.2%           17.4%
3  2145 1    0.68      11.2            2.4%             0.1%             9.5%            2.7%
4  4234 1    0.59      58.7            8.5%             1.3%             8.5%            3.3%
5  2032 2    0.17      57.8           21.7%             0.3%             2.7%            1.2%
6  4171_1    0.07     173.9           58.1%             1.6%             2.1%            0.5%

Checklist:

• Add LatticeOCR
• Handle spanning cells
• Add StreamOCR
• Update docs

[rajasekharponakala]

is PR applicable for non-searchable PDFs?

[vinayak-mehta]

@rajasekharponakala What do you mean by MR?

[rajasekharponakala]

@vinayak-mehta, oops, edited.

[bosd]

Triage bookmark (not closing).

This is a WIP from the previous maintainer that landed promising OCR plumbing but was never finished. The current camelot-dev maintainers have agreed to leave it untouched as a starting point for a future OCR continuation (the alternative — closing as stale — would lose the design discussion and the partially-implemented pdf-to-image + ocrmypdf wiring).

For users who need OCR today, the workaround is the recipe added to docs/user/advanced.rst in #743 — pre-process the PDF with ocrmypdf to add a hidden text layer, then call camelot.read_pdf normally. That covers the common case until this branch is picked up properly.