feat: add ability to extract extra metadata with regex

[MthwRobinson]

Summary

Adds the ability to extract additional metadata by specifying a regex. Also introduces a process_metadata decorator which will allow us to make introduce metadata changes that apply to all document types without having to edit each individual file in the partition directory.

Testing

  from unstructured.partition.text import partition_text

  text = "SPEAKER 1: It is my turn to speak now!"
  elements = partition_text(text=text, regex_match_metadata={"speaker": r"SPEAKER \d{1,3}:"})
  elements[0].metadata.regex_metadata

[qued]

LGTM, a few nits and one question/suggestion. The schema we get back for regex_metadata feels cumbersome (list of dicts for every regex pattern) but I understand why it's that way and can't really think of something better right now.

[cragwolfe]

Neat functionality, but isn't this something someone might want to do post partition_text? I.e., I'm not sure this would always be step zero in a preprocessing pipeline.