[MS] Extend table support for wide tables

[lesyk]

This pull request enhances the handling and extraction of complex tables from PDF files in the markitdown package. It increases the flexibility of the PDF table extraction logic to support documents with a larger number of columns, updates the package version, and adds comprehensive tests for new PDF scenarios. Additionally, it improves repository configuration for handling binary files.

[gagb]

Thanks for the PR @lesyk could you please clarify:

• How were the adaptive constants (0.70 percentile, [25,50] clamp, 10 cols/inch threshold) chosen? Were other values tested?
• Were the existing PDF tests run before and after this change to confirm no regressions?
• Why was the version number bumped?

Can you please also update your description to include commands to run to test your changes and also indicate that you have manually verified all changes, especially if any AI was used to write the code.

[lesyk]

• How were the adaptive constants (0.70 percentile, [25,50] clamp, 10 cols/inch threshold) chosen? Were other values tested?

We have internal testing datasets which has variety of different files After new dataset was added we found that old process of parsing did not work out, thus, making these changes.
As for values, these seem to be more stable from my testing using our datasets.
In previous PRs I have added same synthetic samples, and for each PR add more of them.

• Were the existing PDF tests run before and after this change to confirm no regressions?

I see no regressions on our internal datasets, nor tests I have added previously.

• Why was the version number bumped?

I think I misunderstood versioning for beta channels. I will change to 0.1.5b2. My mistake.

Can you please also update your description to include commands to run to test your changes and also indicate that you have manually verified all changes, especially if any AI was used to write the code.

I am following repos setup: pytest or hatch from root.