Add the default precision for floats in generated GLSL ES shaders
In GLSL ES fragment shaders there is no default precision for floats. The generated shaders are using floats in the return type of a function. These have undefined precision so the shaders should be invalid.
Mesa has started detecting this since 95927c414ff so the tests have started failing.
I’m not sure how I missed this when I did a Jenkins run for that patch :/