Fix html_as_text appearing in every element metadata

[mpolomdeepsense]

This PR fixes the issue described here: Unstructured-IO/unstructured#2463

Now text_as_html will only be available for elements that are HTML strings (contain HTML tags)

E.g. output for non html element

{
    "element_id": "4a44dc15364204a80fe80e9039455cc1",
    "metadata": {
      "coordinates": {
        "layout_height": 3301,
        "layout_width": 2550,
        "points": [
          [170, 13],
          [170, 140],
          [427, 140],
          [427, 13]
        ],
        "system": "PixelSpace"
      },
      "file_directory": "/home/ubuntu/Documents",
      "filename": "purchasing-payment-policy-10.pdf",
      "filetype": "application/pdf",
      "languages": ["eng"],
      "last_modified": "2024-02-02T11:49:38",
      "page_number": 1,
      "parent_id": "e3b0c44298fc1c149afbf4c8996fb924"
    },
    "text": "10",
    "type": "UncategorizedText"
  }

E.g. output for html element

{
    "element_id": "398766f59dd6b37bd38b6d612159cd3e",
    "metadata": {
      "coordinates": {
        "layout_height": 3301,
        "layout_width": 2550,
        "points": [
          [433, 2180],
          [433, 2181],
          [2290, 2181],
          [2290, 2180]
        ],
        "system": "PixelSpace"
      },
      "file_directory": "/home/ubuntu/Documents",
      "filename": "purchasing-payment-policy-10.pdf",
      "filetype": "application/pdf",
      "languages": ["eng"],
      "last_modified": "2024-02-02T11:49:38",
      "page_number": 1,
      "text_as_html": "<table><tbody><tr><td></td><td> Subject Matter Expert / Department</td><td> Contract Review Responsibility</td><td></td></tr><tbody></table>"
    },
    "text": "Subject Matter Expert / Department Contract Review Responsibility",
    "type": "Table"
  }

[yuming-long]

LGTM

[yuming-long]

do we need this here?

[mpolomdeepsense]

It was complaining about the missing text_as_html attribute when I added types to format_table_elements method. Since we use that attribute then I think it should be included in a class. I was also wondering whether it should go inside this class or the parent TextRegion class.

[yuming-long]

I am not following this... is this coming from your test text_as_html_attributes = [fe.text_as_html for fe in formatted_elements] (maybe use if hasattribute(fe, "text_as_html"))?

[mpolomdeepsense]

No, it was coming from models.chipper.UnstructuredChipperModel.format_table_elements method after I added types to the function definition. If there is a reason for not including it in the class then it can stay as was, I will just use the hasattribute like you proposed. I can see that it is used like that in unstructured repo so maybe I should just stick with that? What do you think?

[mpolomdeepsense]

I changed it to hasattr