Columns order for points element shuffled sometimes after parser/deepcopy

[LucaMarconato]

Sometimes (I can't reproduce consistently, it happens rarely), the tests for deepcopy() fail because the deepcopied points have columns with shuffled order. See for instance here: https://github.com/scverse/spatialdata/actions/runs/8281155947/job/22659046336?pr=455

I can't reproduce. I think the problem is in the parser, but since the data is still consistent, I will add a note in the docstring of the parser for points and of the deepcopy for points saying that the order of the columns is sometimes not preserved. This could be also a bug of dask dataframe, and maybe with the newly released dask-expr implementation it will disappear.