Installation check should do retries

[pakrym]

When multiple packages are released from the same pipeline and one depends on another Azure.Core.Experimental -> Azure.Core I tend to kick off both releases at the same time. This causes the Installation Check to fail for Azure.Core.Experimental because Azure.Core is just being published.

I think we should loop and try to restore for some time (30m?) before failing the step.

[benbp]

@ckairen FYI this is related to the ordering you introduced for java. I think I like the eventual approach @pakrym is proposing here better, since it will make releases easier to kick off in the background.

I think we should see if we can find a way to determine whether there is a package we expect to be published before retrying, so we don't stall the release for no reason and make people wait. This might not be possible as I don't think agents have any way to communicate values with each other while running, and additionally I don't know of a way to have a deployment job auto-retry (e.g. schedule to a new agent which would pick up a new variable context).

[benbp]

@pakrym I'm marking this needs-triage just so it comes across for @kurtzeborn, @weshaggard and I in our weekly triage.

[jsquire]

Adjusting tags so that the reports don't yell at me.

[weshaggard]

@benbp is a retry loop similar to what @ckairen did in python enough here, so retry in the install check and not the full job.