Add support for multi-element vector swizzles. Issue #17.

This commit adds support for swizzles like "foo.zy" (if "foo" is,
for example, a float<3> type) as rvalues.  (Still need support for
swizzles as lvalues.)

tests/swizzle-1.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,2,3};
+    float<3> b = a.zxy;
+
+    RET[programIndex] = b.x;
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 3;
+}

tests/swizzle-2.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,10,100};
+    float<3> b = a.zxy;
+
+    RET[programIndex] = b.x + 2*b.y + 3*b.z;
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 132;
+}

tests/swizzle-3.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,10,100};
+    float<6> b = a.zxyyxz;
+
+    RET[programIndex] = b.x + 2*b.y + 3*b.z - 3*b[3] - 2*b[4] - b[5];
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 0;
+}

tests/swizzle-4.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,2,3};
+    float<2> b = a.xy - a.uv;
+
+    RET[programIndex] = b.x + 2*b.y;
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 0;
+}

tests/swizzle-5.ispc

@@ -0,0 +1,15 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,2,3};
+    float<3> b = a.rgb;
+
+    RET[programIndex] = b.x + 2*b.y + 3*b.z;
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 14;
+}

tests/swizzle-6.ispc

@@ -0,0 +1,18 @@
+
+export uniform int width() { return programCount; }
+
+
+export void f_v(uniform float RET[]) {
+    float<3> a = {1,2,3};
+    float<3> b1 = a.yzx;
+    float<3> b2 = b1.yzx;
+    float<3> b3 = b2.yzx;
+    float<3> b = b3 - a;
+
+    RET[programIndex] = b.x + 2*b.y + 3*b.z;
+}
+
+
+export void result(uniform float RET[]) {
+    RET[programIndex] = 0;
+}