Fix bug in &&/|| short-circuiting.

Use full mask, not internal mask when checking "any lanes running"
before evaluating expressions.

Added some more tests to try to cover this case.

tests/short-circuit-12.ispc

@@ -0,0 +1,25 @@
+
+export uniform int width() { return programCount; }
+
+uniform int * uniform ptr;
+
+float foo(uniform float a[]) {
+    int index = (programIndex & 1) * 10000;
+    if (a[programIndex] < 10000 && a[index] == 1)
+        return 1;
+    else
+        return 1234;
+}
+
+export void f_fu(uniform float RET[], uniform float aFOO[], uniform float b) {
+    float a = aFOO[programIndex]; 
+    float a0 = aFOO[0], a1 = aFOO[1];
+    if ((programIndex & 1) == 0)
+        RET[programIndex] = foo(aFOO);
+    else
+        RET[programIndex] = 2;
+}
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = (programIndex & 1) ? 2 : 1;
+}