We want to make sure string users do not depend on the string
implementation. This is the first step.
We need this refactoring *now* to make sure it will not be
super painful to address issue #1175
Before this commit, we were inferring whether an
inductive/structure/class were meta or not. This was bad since the user
had no clue whether the type was trusted (non meta) or not.
Moreover, users could get confused by this behavior and assume the
kernel was allowing trusted code to rely on untrusted one.
