Fix chromsizes in scaling

[Phlya]

Addresses #282 - the chromsizes are extracted in the compute_scaling anyway, so we just need to actually use them.

Fixes the scaling to look like this:

chrom1  start1  end1    chrom2  start2  end2    strand1 strand2 min_dist        max_dist        n_pairs n_bp2
chr1    0       195471971       chr1    0       195471971       +       +       0.0     10.0    0       1954719660.0
chr1    0       195471971       chr1    0       195471971       +       +       10.0    13.0    0       586415880.0
chr1    0       195471971       chr1    0       195471971       +       +       13.0    16.0    1       586415868.0
chr1    0       195471971       chr1    0       195471971       +       +       16.0    20.0    0       781887812.0
chr1    0       195471971       chr1    0       195471971       +       +       20.0    25.0    1       977359744.0
chr1    0       195471971       chr1    0       195471971       +       +       25.0    32.0    55      1368303596.0
chr1    0       195471971       chr1    0       195471971       +       +       32.0    40.0    439     1563775480.0
chr1    0       195471971       chr1    0       195471971       +       +       40.0    50.0    1363    1954719260.0

instead of

chrom1  start1  end1    chrom2  start2  end2    strand1 strand2 min_dist        max_dist        n_pairs n_bp2
chr1    0       -1      chr1    0       -1      +       +       0.0     10.0    0.0     0.0
chr1    0       -1      chr1    0       -1      +       +       10.0    13.0    0.0     0.0
chr1    0       -1      chr1    0       -1      +       +       13.0    16.0    1.0     0.0
chr1    0       -1      chr1    0       -1      +       +       16.0    20.0    0.0     0.0
chr1    0       -1      chr1    0       -1      +       +       20.0    25.0    1.0     0.0
chr1    0       -1      chr1    0       -1      +       +       25.0    32.0    55.0    0.0
chr1    0       -1      chr1    0       -1      +       +       32.0    40.0    439.0   0.0
chr1    0       -1      chr1    0       -1      +       +       40.0    50.0    1363.0  0.0

[Phlya]

I believe the test was actually wrong, now I added explicit removal of (single or double) unmapped pairs, the behaviour before was somehow inconsistent whether the view/regions was provided or not I think. I don't see a reason to include unmapped reads in the scaling calculation, so better to explicitly filter them out IMO.