Investigate reading a subset of BGEN samples (columns)

[tomwhite]

As noted by @eric-czech in #1 (comment), for files with very large numbers of samples we would like a way of just reading a subset of those samples so that we can chunk across samples (columns).

This doesn't seem to be possible in bgen_reader at present (https://github.com/limix/bgen-reader-py/blob/master/bgen_reader/_bgen_file.py#L71), so we should investigate how we could improve this.