Added tests for saturation and some fixes for generic and avx target

tests/paddus_i16.ispc

@@ -0,0 +1,11 @@
+
+export uniform int width() { return programCount; }
+
+export void f_f(uniform float RET[], uniform float aFOO[]) {
+    uniform int16 a = 65535, b = 65535; // max unsigned int16
+    RET[programIndex] = paddus(a, b);
+}
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 65535;
+}