handle empty metadata slot nicely

[d-callan]

ive added a warning, but would we prefer to handle this in the AbundanceData validation and err? it seems a bit heavy handed, but then the warning might be too subtle?

either way i think in practice we wont see this any longer once we are graying the correlation app for studies without metadata, but id still like better logging

[asizemore]

I agree the warning is too subtle. I can't think of a reason to add sample metadata if there is no actual data, so the AbundanceData validation should err i'd say

[d-callan]

In that case it's probably worth a test too..