This allows authors to not have to copy and paste doc comments
from requirements to default implementations to get documentation.
For now, this will only be active if a protocol doesn't inherit
two equivalent requirements from other protocols with doc comments.
Those will need special treatment and it would be better not to just
copy and paste them into the same doc comment. More output structure
is needed to differentiate documentation for different protocols'
requirements.
rdar://problem/16512247
