Enable SelectVariants to handle the * allele properly

[ldgauthier]

I'm sick and tired of dancing around the stupid * allele after I subset a big callset to my truth sample.

If I run SelectVariants with --sample-name sample1 --exclude-non-variants and sample1 is heterozygous ref/* or homozygous for the * then I would like the line in the VCF to get dropped.
My rationale is that the output of that command for less complicated variants is the same as the output if the sample had been run by itself (aside from extra ALTs in the SelectVariants case without --remove-unused-alternates).

I'm willing to concede a --keep-star-only-variants argument to override this behavior.

Any arguments against? Any other nagging * issues for SelectVariants?

[ldgauthier]

Closed via #5129