Fix grouping textlines when bounding box of parent container is wrong

[pietermarsman]

Pull request

Currently, LTLayoutContainer.group_textlines() searches for neighboring text lines within the bounding box of the parent container. If the bounding box of the parent container is wrong, and some of the text lines are outside it, no neighbours are found. Currently, the LTLayoutContainer.group_textlines() skips the text lines where no neighbours are found. Consequently, the content of these lines never appear in the output.

While the PDF is to blame in these cases, we can be more resilient to errorneous PDF's. In this case we should not drop the text line when no neighbouring lines are found. Instead, the line should be returned in its own LTTextBox.

Fixes #381

How Has This Been Tested?

Tested with errorneous PDF from #381. Added unittest for this edge case.

Checklist

• I have added tests that prove my fix is effective or that my feature
  works
• I have added docstrings to newly created methods and classes
• I have optimized the code at least one time after creating the initial
  version
• I have updated the README.md or I am verified that this
  is not necessary
• I have updated the readthedocs documentation or I
  verified that this is not necessary
• I have added a consice human-readable description of the change to
  CHANGELOG.md

[jvalls-axa]

This PR makes all text in figures appear again even if wrong parent container box is used.

Nice catch !