glsl-es-3.10: use highp for int, not float
We need these integers to be highp, otherwise wrapping can occur before 0xffffffff. In fact, the shader doesn't use any floats at all, so let's just reuse the precision statement for float.
Fixes: cacff58c ("glsl-es-3.10: port arb_shader_atomic_counters tests to glsl-es-3.10")