Excel reader does not work for files with just features (no classes) with 3-line header

**What's wrong and how can we reproduce the problem?**

I took the iris data set, removed the class in select rows, and saved it as Excel with "Add type annotations to header".

Loading the file then shows something really strange.

File was written correctly - looks the same as a written file `.tab` file. I guess the empty line causes issues. A file without this Orange's header works well.

Here is the file: [iris.xlsx](https://github.com/user-attachments/files/26838468/iris.xlsx)

Start by loading the attached iris.xlsx through Orange's Excel reader and compare the result with the correctly loaded .tab version. Trace how the reader handles the empty class column in the three-line header; done means feature-only Excel files with that header load without the reported strange output.

Written by the indexing model from the issue text.