This is necessarily incomplete since it can only check ranges that have
a compile-time static size. However, it catches some simple mistakes,
and imposes no runtime cost, so it is still useful.
The minimum supported compiler versions now provide a depth-efficient
implementation of std::make_index_sequence, which removes the last
reason to carry a private implementation. In the case of clang, it
appears to have a special compiler intrinsic used to implement its
std::make_index_sequence.
Switch to the compiler-provided version for both gcc and clang.
