Improve SDK test reliability

[marcpopMSFT]

Creating a tracking item to track all of the efforts to fix our broader test stability issues. Codeflow PRs went from single digit failure rates in September to a quarter of PRs failing now. This is not sustainable.

• 95b6a4a
• Update the wasm install to skip manifest updates #24958
• Allow file changes from tar as it's just a warning #24971
• Fix unstable test build #24592
• Update ubuntu version #24650
  • The runtime occassionally crashes on 18 Ubuntu and they don't plan to resolve given the crash rate is low.
• Next set of three test projects to move to Helix. #23975
  • Helix and non-helix builds run in parallel but are running into each other (see Fix unstable test build #24592 and Allow file changes from tar as it's just a warning #24971 as examples of this). We need to get the last set of tests moved to helix but that's going to be very difficult as dotnet-test has over 1k tests and was never meant to run in helix.
    • Reduce parallel copies of the test .zip file that were overlapping by marcpopMSFT · Pull Request Reduce parallel copies of the test .zip file that were overlapping #25054 · dotnet/sdk (github.com)
    • Next helix test move #25057
    • Next two test projects to move to helix #25197
• SDK tests download over 2k nuget packages. When we run too many PRs in parallel, we get throttled by various feeds that host these packages. We're trying to zip these up so we can pre-download them into the nuget cache.
• Try to lock the mono manifest file to figure out which tests are upda… #24932
  • Something is causing the manifests to get updated into an incoherent state sometimes
• Add back build retry by marcpopMSFT · Pull Request Add back build retry #25141 · dotnet/sdk (github.com)

[marcpopMSFT]

@marek-safar Replying to 95b6a4a, if you see any future failures like that, this is the fix: 95b6a4a. Basically any dotnet new call can potentially run into permissions issues with the template engine hive.

[marcpopMSFT]

@v-wuzhai I've been making a bunch of changes to try to improve the PR failure rate. If you see something not captured here, do let me know. Hopefully it gets better than the really high % failures we've seen in recent months.

[marcpopMSFT]

Adding a few more links on improvements made to our test reliability. That's everything I've been tracking so closing this out for now.
dotnet/installer#14150
#25055
#26843