Instead of grouping every active token on a test expectation as a 'modifier', we should separate those tokens which modify the expected result from those that describe the platform a result is happening on.
<rdar://problem/64913754>