Background:  https://crbug.com/765273#c9 

When optimizing baselines for a test, if it is skipped on a platform, we should consider the platform has a "wildcard" baseline that matches everything, regardless of whether the platform has an existing baseline. This will enable more optimization opportunities. For example in the following scenario:

1. Test X had an old baseline on platform foo
2. Test X is now skipped on platform foo
3. Test X now produces a different result
4. Rebaseline+optimize will generate platform-specific baselines for all platforms but foo, failing to optimize them into a generic baseline because foo is different.