This fixes using samplerviews with explicit depth-only formats set with depth-aspected imageviews, e.g.,.
I'm not 100% sure that this is correct since the spec doesn't explicitly say anything for or against it, but it seems like it should be based on:
- every driver I've tested (ANV/RADV/NV blob) is fine with the behavior
- there's no validation errors
- every other mention of depth-only-aspected formats in the spec refers to only the depth component and format