Fix an implicit truncation warning under MSVC 32-bit

Since the return value of `NumBits()` will always fit into `size_t`,
use an explicit cast to silence the implicit conversion warning.

Fixes #1384

PiperOrigin-RevId: 518041598
Change-Id: If2f2456db4b27b78e9ea9e026dce610953bd5bfb

absl/random/internal/fast_uniform_bits.h

@@ -57,9 +57,10 @@ constexpr UIntType IntegerLog2(UIntType n) {
 // `PowerOfTwoVariate(urbg)`.
 // `PowerOfTwoVariate(urbg)`.
 template <typename URBG>
 template <typename URBG>
 constexpr size_t NumBits() {
 constexpr size_t NumBits() {
-  return RangeSize<URBG>() == 0
+  return static_cast<size_t>(
+      RangeSize<URBG>() == 0
           ? std::numeric_limits<typename URBG::result_type>::digits
           ? std::numeric_limits<typename URBG::result_type>::digits
-             : IntegerLog2(RangeSize<URBG>());
+          : IntegerLog2(RangeSize<URBG>()));
 }
 }
 
 
 // Given a shift value `n`, constructs a mask with exactly the low `n` bits set.
 // Given a shift value `n`, constructs a mask with exactly the low `n` bits set.