Phenopackets do not allow for representation of provenance of key data elements

[bpow]

The ability to represent per-element provenance of things like VariantInterpretation and Diagnosis are limiting with respect to how we've considered using phenopackets to store/transmit ClinGen-curated evidence.

For example, I am hesitant to include a VariantInterpretation which a ClinGen biocurator has abstracted from the literature because, in the absence of proper attribution the presence of this data in a ClinGen-produced phenopacket could easily be misinterpreted as ClinGen making an assertion of the pathogenicity of this variant.

Other elements that represent "judgement" could also arguably need the ability to represent provenance/attribution-- for instance, GenomicInterpretation, Diagnosis.

[bpow]

I'll note that this is related to #441 in that the classification and determination of whether or not a diagnosis is established are two separate assertions that may be made by different people.

Arguably, even observations of presence or absence of a clinical feature (as HPO term) is an assertion that could benefit from some metadata. For example, in the clinical world, a "nonfocal neuro exam" means different things to me depending on whether it was performed by a neurologist I trust :). Phenotypic observations do have the ability to provide Evidence in phenopackets, what I'm suggesting here is that we allow for Evidence objects (which permit a uri ExternalReference) to be applied to things like Interpreation, GenomicInterpreation and VariantInterpretation.