Fixes to the new solver & VersionOrder

[mcg1969]

Fixes motivated by several issues reported by testers: #2033, #1912, #1476, #1334, etc.

The key problem with regards to features was that the pruning pass was eliminating features from consideration prematurely. We've moved the elimination pass until after the first scan to make sure to catch all of the track_features entries that would be caught by the specs.

I've also adjusted the optimization passes so that they proceed in the following order:

1. Choose the latest possible versions of the explicitly requested packages. In this pass, we don't try to break ties with build numbers or build strings.
2. Minimize the number of track_features selected to achieve 'conda update anaconda' should fail when no internet is available #1. This won't touch track_features that are embedded in installed metapackages or packages like python that always have them. But it will ensure that no additional, unnecessary track_feature values are added to the mix.
3. Now choose the latest possible versions of all explicitly requested packages and their dependencies. At this stage, we do break ties between packages as follows:
   • Packages with more valid features are preferred over packages with fewer.
   • If that doesn't break a tie, then packages with a larger build number are preferred over packages with a smaller build number.

If anyone is nerdy enough (like me) to parse this and consider its consequences, please do, and let me know if you have any concerns.

[github-actions]

Hi there, thank you for your contribution to Conda!

This pull request has been automatically locked since it has not had recent activity after it was closed.

Please open a new issue or pull request if needed.