Account for rowspan in table headingColumns

[Witoso]

📝 Provide a description of the improvement

Improvement from @bendemboski.

When computing headingColumns during upcasting, the logic wasn't factoring in s with rowspan attributes that "add" a heading column to subsequent rows. See tests and comments in the code for more detail.

---

If you'd like to see this improvement implemented, add a 👍 reaction to this post.

[wimleers]

Tracking this on drupal.org too: https://www.drupal.org/project/drupal/issues/3395434

[Timmy-Cos]

I notice this issue seems very similar to the recently solved #17556, and can be reproduced in a similar way:

1. Open for example https://ckeditor.com/docs/ckeditor5/latest/examples/builds-custom/full-featured-editor.html
2. Create a 3x3 table
3. Merge the first cells of the second and third row
4. Set the first two columns as headers
5. editor.setData(editor.getData()) in the console will then strip a header column each time it is run

Alternatively, just open that page and run
editor.setData('<figure class="table"><table><tbody><tr><th>th cell</th><th>th cell</th><td>td cell</td></tr><tr><th rowspan="2">rowspan th cells</th><th>th cell</th><td>td cell</td></tr><tr><th>th cell</th><td>td cell</td></tr></tbody></table></figure>')
The resulting table only has one header column instead of two. It appears the issue lies with setData rather than getData in this case.

✔️ Expected result
editor.setData() returns a valid table with all the provided header columns

❌ Actual result
editor.setData() returns a table with missing header columns

Like #17556 I believe this issue is a regression introduced in 37.0.0-alpha.1. @Witoso would it be fair to relabel this as a bug? Would it be possible for the attached PR to get another look in?