new edge case for balancing with ignoring diagonals

[sergpolly]

when balancing genome-wide and invoking ignore_diags (which is 2 by default) cooler will ignore neighboring trans data as well ... in case there are tiny-tiny chromosomes, e.g. 1-pixel in size:

if this is my hi-c heatmap:

this is what's going to ignored, when balancing with ignore_diags=2 :

note that the pixel related to chr2-chr3 trans-interactions is also ignored ...

1. case for tiny crhomosomes : - maybe there are tiny contigs , maybe some sorta crispr constructs, plasmids, - they could be tiny and maybe represented as a standalone chromosome ...
2. is balancing justified for such single-bin chromosomes, - don't know - probably not
3. is it a desired behavior or should cis- only diagonal be ignored ? probably cis-only - as with ignore-diags we are trying to avoid Hi-C artefacts ...
4. maybe there is no clearly right or wrong solution, then it's just important to document stuff like that