Use chromsizes from the pairs header in scaling (fixes empty n_bp2)

[OfficialBishal]

Closes #282.

Running pairtools scaling without --view left the n_bp2 column empty: compute_scaling reads the chromsizes from the .pairs header (via read_pairs) but discarded them (pairs_df, _, _ = ...), so bins_pairs_by_distance fell back to region ends of -1 and n_bp2 could not be computed.

This uses the header chromsizes when the caller doesn't pass an explicit chromsizes, matching the existing intent — the CLI passes the stream "so that it can parse the header automatically". An explicit chromsizes argument and the DataFrame input path are unaffected (the change is gated on chromsizes is None).

On tests/data/mock.pairsam (header chr* = 100):

• before: end1 == -1 for every bin, cis n_bp2 all 0, trans n_bp2 garbage.
• after: end1 is the chromosome length from the header, and n_bp2 is populated.

Added test_scaling_uses_header_chromsizes covering the no-view case.

[Phlya]

I think I already fixed it here: #283

[OfficialBishal]

Ah, thanks @Phlya — #283 already fixes this (and handles the unmapped pairs too). Closing in favor of yours. If the small regression test I added here is useful, happy to send it over to #283. Sorry for the duplicate!