Improve baseline determination when comparing VMR and MSFT builds

[mmitche]

Today, when we do asset comparison between VMR and MSFT builds, we find artifacts with the same version as the VMR produced artifact and compare the contents. Unfortunately, this will cause us to miss comparisons. Artifacts not produced by the VMR will not be validated. This issue covers revamping this process to determine a proper baseline.

I think this could be done in a few ways:

• Manually prepared baselines that are updated on a cadence.
• Use gather drop, plus some heuristics to download a baseline.

Any way this is done, the goals are:

• We should be able to tell what artifacts are missing from the VMR build.
• The baseline should be relatively simple to update
• The baseline should be coherent.
• The baseline should not rely on 1:1 version matches

[mmitche]

@wtgodbe This is the other issue for you to focus on outside of the hosting bundle. Figuring out how we do our assets comparisons in a robust manner.