Don't add AOT package in restore for < net7

[jtschuster]

Fixes #30814.

In order to avoid errors for invalid RIDs for AOT during restore, we need to avoid trying to restore a package for ILC when TFM < 7. Then, we need to warn during publish when there are no ILC packages found during restore to make sure we don't still go through with publish with PublishAOT=true and TFM<7.

[sbomer]

Could you check if the other publish scenarios have the same problem? I think you could get it with PublishReadyToRun (maybe you need to opt into crossgen2) on downlevel TFMs too.

[jtschuster]

I forgot to follow up on the PR feedback and let this get stale. The issue was still present in preview 4, and this fix is still necessary. @sbomer @agocke Can you take another look?

[sbomer]

Thanks for following up on this!

Looking at this a second time, I don't think adding a TFM condition to the task is a good idea. The source of truth for which TFMs are supported by the various scenarios (not just AOT) lives in https://github.com/dotnet/installer/blob/main/src/redist/targets/GenerateBundledVersions.targets, and ProcessFrameworkReferences is the current choke point that produces errors for all of these scenarios. I've been leaning on this to improve the error messages (see #32943 and #33041).

The current restore behavior (restoring for all TFMs) is by design per #30814 (comment), so if we're unhappy with this behavior I think we should be looking at a broader solution that changes the design.