Silent deletion of data should be construed as a bug, not the default behavior

[D3SL]

I just ran into the issue from #208 with someone I'm helping learn python. In fact that's actually the only reason we realised anything was wrong, because it was a small human readable practice dataset they knew some of the columns shouldn't have been completely empty. There were no warnings, no error messages, just data going missing completely silently. Nothing in Polars' breaking changes documentation suggested this was a possibility, and nothing in Fastexcel's documentation warns of possible silent data loss either.

Imho inferring a "Null" column should be treated as a failure state. A user should need to explicitly opt-in via argument to allow the possible loss of data with a warning. Deleting an entire column of data completely silently because an excel spreadsheet just happened to have been sorted in a particular order is a showstopping issue. I can not state in strong enough terms how much I disagree with the choice to have this as a default behavior.