Attach per-haplotype tip labels from the input VCF

[wnsplim]

Currently SINGER reads VCF genotypes positionally and discards the sample names, so the output has no tip labels (samples come out enumerated 0…n-1). This makes mapping results back to samples awkward for some downstream analyses.

We can re-attaches names from the input VCF when converting SINGER output to tskit: each diploid sample i (in VCF column order) becomes individual i with label set to the VCF sample name, and the two tips are named [sample]_0 & [sample]_1.

Edits:

• merge_ARG.py: new optional --vcf flag (off by default; behavior unchanged unless passed).
• convert_long_ARG.py: uses the already-required -vcf prefix, so labeling happens automatically. (It now reads {prefix}.vcf, which is present in the normal workflow)

If these should be handled differently, this PR can be dropped.