#include "tensorstore/util/int2.h"
template <typename T>
  requires std::is_convertible_v<T, int8_t>
explicit constexpr tensorstore::Int2Padded::Int2Padded(T x);

Possibly lossy conversion from any type convertible to int8_t.