Option to treat float64 as float32 in write_vcf

[timothymillar]

The write_vcf function doesn't support float64 which makes sense given it's not supported by the VCF specification. However, manually converting float64 arrays to float32 is a bit of a pain, so I wonder if we could have a flag to treat float64 as float32? It's also a little inconsistent as we allow all integer types where the VCF spec states that the "Integer" type is int32 (code).