Rudder

In some cases, especially in geographically distributed infrastructures, even more when machine are created from golden imagine, several machine can have the same hostname.

For now, we refuse to accept a new node when an already accepted node with the same hostname exist. This is a sane default and it should be kept.
In parallel, we should add an option to accept node in that case.

Some things will broke: at least remote run is based on hostname.
None the less, we must ensure that rudder main usage is still OK (grouping, policy generation and reports) - and it should since we are only based on node ID for these things.

In parallel, we must make that use case a common one for a next major, and work to add:

- remote run based on something else than hostname (uuid would be best),
- node display name/description customization (so that different names are displayed differently in tables)
- anything that we discover