Expose AVX2 instruciton

Currently we do not test AVX2 drawing code paths.

AVX2 along with FMA was introduced with the Haswell architecture[3] (fourth
generation Intel Core, according to the marketing name), first launched in
late 2013. That makes the architecture over 5 years old and it means a lot of
our users have the feature.

The current setup uses the same CPU for all integrations. it means once the feature is added there is other code path not tested. There are two options:
1. Decide what super set of CPU features we want to have to test Qt on (globally, for all branches). In places in which it matters we can disable some features through QT_NO_CPU_FEATURE to test multiple code paths.
2. Add a way to inject the CPU features in the same way as configure features.

Obviously the 2nd one is better, but it also requires some code in Coin and changes in OpenNebula configurations. So maybe in meanwhile 1st can be used.