Use @llvm.readcyclecounter to implement stdlib clock() function.

Also added a test for the clock builtin.

tests/clock.ispc

@@ -0,0 +1,14 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_f(uniform float RET[], uniform float aFOO[]) {
+  unsigned uniform int64 a = clock();
+  float x = pow(sqrt(aFOO[programIndex]), 5.5);
+  unsigned uniform int64 b = clock();
+  RET[programIndex] = (b - a) > 0 ? 1 : 0;
+}
+
+export void result(uniform float RET[]) {
+  RET[programIndex] = 1;
+}