Flakiness caused by lack of retries

[davidfowl]

Seems like we don't retry downloads which can load to flakyness when they fail (and they do quite often see #20561 (comment)).

@dougbu also thinks this is a general problem with the scripts that we have under eng/*.

[dougbu]

It's not just tests that are hurt here. A couple of the Windows jobs in https://dev.azure.com/dnceng/public/_build/results?buildId=590331 failed almost immediately due to https://nodejs.org flakiness.

[dougbu]

Some common patterns to look at are in eng/common/tools.ps1|sh Need to look at any scripts that run on CI or Helix (test) agents, probably most of what's in eng/scripts (though it's fine to ignore the VS installation stuff).

[dougbu]

Not completely done yet 😺 @HaoK's work in #22898 covers the Windows situation. Still need to add something like --retries 5 to curl commands in our *.sh files.