Fix wrong denominator in max_missing_an fraction filter (closes #998)

[rehanxt5]

What

When max_missing_an is passed as a float, the missingness fraction was computed by dividing by an (called alleles) instead of total alleles. This made the filter consistently more restrictive than intended.

Why it's wrong

A SNP with exactly 5% missing data in 100 diploid samples (10 missing out of 200 total) would compute 10 / 190 = 0.0526 and fail a max_missing_an=0.05 threshold — even though 5% ≤ 5% should pass. At high missingness levels the formula produces values greater than 1.0, which is not a valid fraction.

Change

Before:

an_missing = (ds_out.sizes["samples"] * ds_out.sizes["ploidy"]) - an
if isinstance(max_missing_an, float):
    an_missing_frac = an_missing / an  # wrong: called alleles as denominator
    loc_missing = an_missing_frac <= max_missing_an

After:

an_total = ds_out.sizes["samples"] * ds_out.sizes["ploidy"]
an_missing = an_total - an
if isinstance(max_missing_an, float):
    an_missing_frac = an_missing / an_total  # correct: total alleles as denominator
    loc_missing = an_missing_frac <= max_missing_an

I also updated the tests for tests/anoph/test_snp_data.py

Notes

• The integer path (max_missing_an as int) is unaffected.
• The min_minor_ac float path below this uses an (called alleles) as denominator which is correct for allele frequency — that path is unchanged.

closes #998

[rehanxt5]

@jonbrenas hey 👋 can you please review this