Fix issue #2: use zero extend to convert bool->int, not sign extend.

This way, we match C/C++ in that casting a bool to an int gives either the value
zero or the value one.  There is a new stdlib function int sign_extend(bool)
that does sign extension for cases where that's desired.

tests/movmsk-opt.ispc

@@ -20,7 +20,7 @@ export void f_fu(uniform float RET[], uniform float aFOO[], uniform float b) {
 
     float v = float4(1,1,0,0);
     bool b = (v == 1.);
-    ret = __movmsk(((int)b));
+    ret = __movmsk((sign_extend(b)));
     RET[programIndex] = ret;
 }
 

tests/sign-extend-1.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_f(uniform float RET[], uniform float aFOO[]) {
+    float a = aFOO[programIndex]; 
+    bool b = (a == 1.);
+    int32 s = sign_extend(b);
+    RET[programIndex] = (s == 0xffffffff) ? 16 : 2;
+}
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 2;
+    RET[0] = 16;
+}

tests/sign-extend.ispc

@@ -0,0 +1,14 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_f(uniform float RET[], uniform float aFOO[]) {
+    float a = aFOO[programIndex]; 
+    bool b = (a == 1.);
+    RET[programIndex] = (int)b;
+}
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 0;
+    RET[0] = 1;
+}