Background:

webkit-patch rebaseline-cl tries new baselines to download from try jobs and sometimes results can't be fetched for all platforms.

If we're downloading new baselines for a test, and if we don't have all results for all platforms, there are two possibilities:
 1. The platform missing results has the same results as other platforms
 2. There are platform-specific baselines

In case (1), we want to use results from another platform, and in case (2) we just don't want to download any results for the platforms without results.

Proposal (idea from Walter):

> Scrutinize the baselines involved and inform the developer whether (before their change), for example, the to-be-filled baselines differed from the platform(s) with which we would fill them if --fill-missing is used. I think we have all of the info we need to determine that. But it could be a pain Git-command-wise to scrutinize things, or just the add'l code complexity to do this may not warrant the benefit. Also, it's always possible that the developer's new change introduces differences that weren't there before.... It would at least give the user some signal re: likelihood of using --fill-missing doing the right thing vs. obviously breaking.

(Fredrik, do you have any thoughts about whether this would have been helpful?)

That is, if there are currently platform-specific baselines for the given platform/test, then --fill-missing might be dangerous and the tool could print a warning, and If there are no platform-specific baselines, then it's probably OK, and --fill-missing could probably be the default behavior.