[ancestry] Use assembly when choosing which model to use

[akotlar]

Background:

A concise description of why we're pursuing this task.

Currently we assume all ancestry requests are for hg38. As we don't have a liftover step, preferring to natively process hg19 vcf's, we need to carry the assembly in the ancestry submission job request, so that the appropriate model can be loaded.

In #421 we thread through AncestryData.

In this ticket we need to update ancestry/inference.py to load the appropriate model based on the AncestryData.assembly

Completion Criteria:

What is the goal, and how do we know this task is accomplished?

In inference.py AncestryData.assembly is used to choose the model to load

Implementation Summary

Give enough technical detail for a teammate to understand the proposed solution.

AncestryJobData needs to now expect an assembly property (string valued). This property then needs to be threaded through to the ancestry inference procedure, which currently hardcodes an hg38 expectation