Add a test for floating point exponent overflow

This adds a test with a floating point exponent that can't fit into a
32-bit int.

Test/positive_infinity.frag

@@ -0,0 +1,11 @@
+#version 300 es
+precision highp float;
+out vec4 my_FragColor;
+void main()
+{
+    // Out-of-range floats should overflow to infinity
+    // GLSL ES 3.00.6 section 4.1.4 Floats:
+    // "If the value of the floating point number is too large (small) to be stored as a single precision value, it is converted to positive (negative) infinity"
+    float correct = isinf(1.0e2147483649) ? 1.0 : 0.0;
+    my_FragColor = vec4(0.0, correct, 0.0, 1.0);
+}