Consider allowing duplicate column names in reading ASCII tables

[taldcroft]

From discussions in #5373, it might be reasonable to allow duplicate column names in an input table and automatically rename as <colname>_<N>, where N is incremented until the column name is unique.

np.recfromtext uses the _<N> convention while pandas does .<N>. Both packages seem to have buggy / limited implementations in terms of handling conflicts that arise from this auto-renaming.

Implementation effort is not clear at this point.