Develop layout analysis R&D pipeline

Developed a pipeline in the pdf-parsing-evaluation repo. This pipeline integrates:

classes for a document image, bounding box labels for an image, and a bounding box label dataset (multiple pages)
multiple possible pre-processing and post-processing steps (although none implemented so far)
the ability to add a model, and functions to easily add one from Layout Parser
the ability to map labels names using a dictionary mapping, e.g. when comparing models trained on datasets with different label schemes
model evaluation: AP and AR at a fixed IOU or a range of IOUs, at a certain confidence threshold.

climatepolicyradar / navigator