Fixup the chromosome zarr encoding bug

[ravwojdyla]

If we start the inference from the beginning (i.e. chr1) we encode chr_name as <U4, but at some point <U5 is required.

Below is the list of all the chr_name:

{'chr1', 'chrX', 'chr13', 'chr16', 'chr4', 'chr7', 'chr14', 'chr11', 'chr22', 'chr20', 'chr17', 'chr2', 'chr15', 'chr10', 'chr19', 'chr5', 'chr9', 'chr21', 'chr12', 'chr8', 'chr6', 'chr18', 'chr3'}

@yonromai I originally thought it was the token character, but it's actually the chromosome name, which is a lot more useful downstream, so we should keep it :)