fix: convert to pixels

[qued]

Fixes a bug where coordinates in point-coordinates were not being converted to pixel-coordinates in agreement with the specified DPI.

Testing:

Parsing layout-parser-fast.pdf Should properly extract the text from the title block.

[cragwolfe]

works for me in an unstructured project:

$ PYTHONPATH=. python -c "from unstructured.partition.pdf import partition_pdf; elems = partition_pdf('example-docs/layout-parser-paper-fast.pdf'); from rich.pretty import pprint; pprint(elems[0].to_dict()); pprint(elems[7].to_dict())"
{
│   'element_id': '015301d4f56aa4b20ec10ac889d2343f',
│   'coordinates': (
│   │   (438.1653747558594, 312.04864501953125),
│   │   (438.1653747558594, 413.18951416015625),
│   │   (1274.8990478515625, 413.18951416015625),
│   │   (1274.8990478515625, 312.04864501953125)
│   ),
│   'text': 'LayoutParser: A Unified Toolkit for Deep Learning Based Document Image Analysis',
│   'type': 'Title',
│   'metadata': {
│   │   'filename': 'example-docs/layout-parser-paper-fast.pdf',
│   │   'page_number': 1
│   }
}
{
│   'element_id': '44b6b64d159ac94e2f6910fd5055a68f',
│   'coordinates': (
│   │   (574.2962646484375, 557.7698364257812),
│   │   (574.2962646484375, 870.4000244140625),
│   │   (1111.1734619140625, 870.4000244140625),
│   │   (1111.1734619140625, 557.7698364257812)
│   ),
│   'text': 'Universityo',
│   'type': 'ListItem',
│   'metadata': {
│   │   'filename': 'example-docs/layout-parser-paper-fast.pdf',
│   │   'page_number': 1
│   }
}