Description of problem:
When touching an existing bridge, any existing attached port, even if unmanaged, is implicitly taken over by nmstate/NM and a relevant profile in NM is created for it.
This behavior is not working well in a virtualized environment: vNIC/s from VM/s are attached and detached from the bridge by the hypervisor (e.g. libvirt) and such interfaces are not intended to be managed by nmstate (or NM). However, with the current behavior, any editation through nmstate to the bridge will cause the existing tap devices to be managed by NM.

Currently, state=DOWN is associated with an unmanaged device (something which should probably be fixed as well), therefore, for bridges, interfaces which are reported as `DOWN` should not be implicitly managed by nmstate/NM.

Version-Release number of selected component (if applicable):


How reproducible:
100%

Steps to Reproduce:
1. Define a linux-bridge using nmstate.
2. Create a dummy or tap device and attach it to the bridge using the `ip` tool.
3. Create a 2nd dummy device and attach it to the bridge using nmstate.

Actual results:

The 1st device has now a profile and is fully managed by NM/nmstate.

Expected results:

The first device should not be managed by NM/nmstate and should appear as `DOWN`.

Additional info: