Correct C FixedMul() off-by-one errors

The FixedMul() C implementation would produce off by one results,
causing constant desyncs on 64 bit builds and builds without an
ASM implementation of the function.

This is fixed by shifting instead of dividing, possibly avoiding
rounding errors.

src/m_fixed.c

@@ -33,7 +33,9 @@
 */
 fixed_t FixedMul(fixed_t a, fixed_t b)
 {
-	return (fixed_t)((((INT64)a * b) ) / FRACUNIT);
+	// Need to cast to unsigned before shifting to avoid undefined behaviour
+	// for negative integers
+	return (fixed_t)(((UINT64)((INT64)a * b)) >> FRACBITS);
 }
 
 #endif //__USE_C_FIXEDMUL__