wordacc: text stream is too long

When running `wordacc` on large files in Devanagari script I am getting the error `wordacc: text stream is too long`.

`accuracy` program from this repo works fine on the same input files, though the one from earlier version was getting same error - see https://github.com/ryanfb/ancientgreekocr-ocr-evaluation-tools/issues/2

Example files:
* [wordacc report - size o](
https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/report_allfontsMinus_0.891_170479_475200-allfonts-ALL-wordacc.txt)
* [accuracy report](https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/report_allfontsMinus_0.891_170479_475200-allfonts-ALL.txt
)
* [OCRed text - 1.45 MB](https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/ocr-allfontsMinus_0.891_170479_475200-allfonts-ALL.txt)
* [Ground Truth](https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/gt-allfonts-ALL.txt)

* [Eval using ocrevalUAtion](https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/report_allfontsMinus_0.891_170479_475200-allfonts-ALL.html)

It works when used with smaller files, example [wordacc report](https://github.com/Shreeshrii/tesstrain-deva/blob/master/reports/allfonts-eval-allfontsMinus_0.891_170479_475200/report_allfontsMinus_0.891_170479_475200-allfonts-Adobe_Devanagari-wordacc.txt)