Reliability issues with some of the automated tests

[bobbrow]

• Add some mechanism to deal with tests that have known random failures, such as having alternative baselines, or a retry count for tests that have known random failures. But it would be good to not completely bury the failures (which could be real bugs).
• recursive_includes test has a random failure on Linux (potentially a real race condition bug).
• fix custom configuration test failure.
• fix squiggles test failure.
• fix intellisense_clients random failure on shutdown
• multiroot/SymbolSearch randomly gives no results.
• multroot/FindAllReferences randomly misses results.
• multiroot/SetFolderSettings randomly gets stuck.
• multiple tests/OpenFile: IntelliSense process randomly shuts down.
• database_clearing test random gets stuck on the Linux/Mac build machines -- doesn't appear to be worth investigating because it doesn't repro on my local machines.

...some of these may have gotten fixed, I haven't checked recently.

[bobbrow]

Ideally, this would not be needed, but we could update the affected test(s) so that they are more stable.

[sean-mcmanus]

UPDATE: The multiroot test on Windows is still failing and the fallback/timeout-retry mechanism has some failure I need to investigate.

[sean-mcmanus]

This issue somewhat related to #6307 (debug build tests).