Gix GCC cast ambiguity.

GCC complains about ambiguous casting mxfloat specifically to float8_e5m2fnuz.
The cause seems to be missing constructors.  Here we make the constructor
similar to all other float8 types, which eliminates the ambiguity.

PiperOrigin-RevId: 721882407

ml_dtypes/include/float8.h

@@ -399,17 +399,8 @@ class float8_e5m2fnuz : public float8_base<float8_e5m2fnuz> {
   using Base::Base;
 
  public:
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e5m2& f8)
-      : float8_e5m2fnuz(ConvertFrom(f8)) {}
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e4m3& f8)
-      : float8_e5m2fnuz(ConvertFrom(f8)) {}
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e3m4& f8)
-      : float8_e5m2fnuz(ConvertFrom(f8)) {}
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e4m3b11fnuz& f8)
-      : float8_e5m2fnuz(ConvertFrom(f8)) {}
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e4m3fn& f8)
-      : float8_e5m2fnuz(ConvertFrom(f8)) {}
-  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(const float8_e4m3fnuz& f8)
+  template <typename T, RequiresIsDerivedFromFloat8Base<T> = 0>
+  explicit EIGEN_DEVICE_FUNC float8_e5m2fnuz(T f8)
       : float8_e5m2fnuz(ConvertFrom(f8)) {}
 
   constexpr float8_e5m2fnuz operator-() const {