OpenCV findContours for table detection

See

https://nanonets.com/blog/table-extraction-deep-learning/#code-traditional-approaches

This can be added as an alternative near to substitute get_positions function.