Explicitly static cast to long double (#267)

This fixes a very verbose warning that I noticed when installing the
library in a Visual Studio project on Windows.  And it just seems like a
clearly good change for its own sake, anyway.

au/code/au/magnitude.hh

@@ -378,7 +378,7 @@ constexpr MagRepresentationOrError<T> root(T x, std::uintmax_t n) {
     // Always use `long double` for intermediate computations.  We don't ever expect people to be
     // calling this at runtime, so we want maximum accuracy.
     long double lo = 1.0;
-    long double hi = x;
+    long double hi = static_cast<long double>(x);
 
     // Do a binary search to find the closest value such that `checked_int_pow` recovers the input.
     //