Improve performance of column name matching

This is horribly slow when ~20k columns are specified

https://git.fmrib.ox.ac.uk/fsl/funpack/-/blob/master/funpack/importing/filter.py#L174