Improve parsed-price accuracy

[EtienneLamoureux]

Situation

Prices are prefixed with the ¤ symbol. This symbol is not in the english training set of Tesseract and is read as a random character. When this character is read as a digit, it inflates the prices read by an order of magnitude, i.e. ¤900 becomes 2900.

Tasks

1. Experiment with heuristics to mitigate the issue
   1. Thousands are always separated by a comma , and groups of digit are only up to 3 long
   2. Only 1 digit is present before the comma , when the price is listed in kilo units K
   3. Others

Results

1. The ¤ character doesn't inflate prices