Openstack provider clean-up

This issue is to do some clean-up in the openstack provider logic, that in some cases is not very clear or optimal. A few examples:

In order to get a network by name or ID we iterate through all networks instead of getting it directly with the proper API call:
https://github.com/kubermatic/machine-controller/blob/master/pkg/cloudprovider/provider/openstack/helper.go#L217

Same applies for security groups:
https://github.com/kubermatic/machine-controller/blob/master/pkg/cloudprovider/provider/openstack/helper.go#L184

The logic we use to get the default subset is a bit obscure: if we have just one subnet in the network provided in input we return it otherwise we obtain the subnets from the API and we return the first if any or we throw an error. If there is a reason behind that, I have some doubts, we should at least comment it properly.
https://github.com/kubermatic/machine-controller/blob/master/pkg/cloudprovider/provider/openstack/helper.go#L416

Start in pkg/cloudprovider/provider/openstack/helper.go at the network lookup around line 217, security-group lookup around line 184, and default-subnet logic around line 416. Trace the OpenStack API usage and the reasoning behind the subnet behavior. Done means the lookup paths are clear and efficient, and the default-subnet behavior is either justified in comments or made consistent with the intended provider behavior.

Written by the indexing model from the issue text.