build(deps): split up dependencies by document type

[MthwRobinson]

Summary

Closes #611. Updates the dependencies to create a slim base installation + extras for each document type. The all-docs extra includes the dependencies for all doc types. The local-inference extra still remains as an alias for all-docs since that's the extra many currently users use to install all of the dependencies.

As a follow on, the PDF and image requirements could be broken down further by strategy.

Testing

• In a fresh virtual environment, run make install-base and pip install -r requirements/test.txt and pip install -e .. The following tests should pass:
  • pytest test_unstructured/partition/test_text.py
  • pytest test_unstructured/partition/test_email.py
  • pytest test_unstructured/partition/test_html_partition.py
  • pytest test_unstructured/partition/test_text.py
  • pytest test_unstructured/partition/test_xml_partition.py
• Run make install-csv. The following tests should pass:
  • pytest test_unstructured/partition/test_csv.py
  • pytest test_unstructured/partition/test_tsv.py
• Run make install-docx. The following tests should pass:
  • pytest test_unstructured/partition/test_doc.py
  • pytest test_unstructured/partition/test_docx.py
• Run make install-odt. The following tests should pass:
  • pytest test_unstructured/partition/test_odt.py
• Run make install-pypandoc. The following tests should pass:
  • pytest test_unstructured/partition/test_epub.py
  • pytest test_unstructured/partition/test_org.py
  • pytest test_unstructured/partition/test_rst.py
  • pytest test_unstructured/partition/test_rtf.py
• Run make install-markdown. The following tests should pass:
  • pytest test_unstructured/partition/test_md.py
• Run make install-msg. The following tests should pass:
  • pytest test_unstructured/partition/test_msg.py
• Run make install-pdf-image. The following tests should pass:
  • pytest test_unstructured/partition/test_pdf.py
  • pytest test_unstructured/partition/test_image.py
• Run make install-pptx. The following tests should pass:
  • pytest test_unstructured/partition/test_ppt.py
  • pytest test_unstructured/partition/test_pptx.py
• Run make install-xlsx. The following tests should pass:
  • pytest test_unstructured/partition/test_xlsx.py
• Run make install-ci and pip install argilla weaviate-client. make test should pass

Test test_unstructured/partition/test_json.py is currently a little weird because there are tests in there that use partition, making it hard to test partition_json isolation. I opened #993 to address that.

[cragwolfe]

The pattern looks good to me. Hoping to merge #997 ahead of this to get the field name change released first.

[qued]

LGTM, a few comments and suggestions.