Feat: improve image extraction by supporting all types of image elements detected by detection models

[christinestraub]

Closes #285.

Summary

• support extracting elements with types Picture and Figure
• add a class ElementType for the element type constants and use the constants to replace element type strings

Testing

PDF: algebra-graph-level1-1.pdf

from unstructured_inference.inference.layout import DocumentLayout

doc = DocumentLayout.from_file(
    filename="algebra-graph-level1-1.pdf",
    extract_images_in_pdf=True,
)

[benjats07]

LGTM
In the future we may explore the possiblity of just renaming Picture and Figure to be Image and simplify this.