Go/No go decision for instantiate to bounds as implemented

[leafpetersen]

We have an initial implementation of the proposed instantiate to bounds semantics, as discussed here: #27526 .

The language team needs to make a final decision about whether the impact of these changes is reasonable. I have CLs prepared to address most of the changes required to internal code. I will follow up on several issues with the language team.

Subsequently I will follow up here with a decision, either to:

1. roll back the initial implementation and revisit for next release
2. relax the restriction to the minimum that we are sure that we want
3. roll forward with the implementation and land all of the CLs to fix internal code

cc @kevmoo @mit-mit @floitschG @munificent @eernstg @lrhn

[leafpetersen]

We will relax the restriction slightly as described here: #28580

Otherwise we're ok with the impacts of this change.