in which the private/restricted-in-public error messaging gets specific
April 2016's Issue #33174 called out the E0446 diagnostics as
confusing. While adding the name of the restricted type to the message
(
548e681f) clarified matters somewhat, Esteban Küber pointed out that we
could stand to place a secondary span on the restricted type.
Here, we differentiate between crate-visible, truly private, and
otherwise restricted types, and place a secondary span specifically on
the visibility modifier of the restricted type's declaration (which we
can do now that HIR visibilities have spans!).
At long last, this resolves #33174.