relay: send stats to portal

[thomaseizinger]

The portal needs to know the utilisation of a relay in order to load-balance them for new connections.

The critical stats here are:

• Relative utilisation of allocations (available port range / used port range)
• Relative CPU utilisation of the entire machine (/proc/stat?)

Other related metrics are:

• Active channel bindings: Those are per peer, so more difficult to report and not as useful for load balancing because the conclusion would have to be "you can still use this relay, just not for this one client / gateway". We have lots of channels per peer and we should do snownet: Check if direct connection is possible again after some time X #4164 to be more considerate with this resource.
• Throughout: Not really useful for load balancing because we don't know the max. possible throughput.

[AndrewDryga]

We might use the same metrics for #4415.

[thomaseizinger]

The portal needs to know the utilisation of a relay in order to load-balance them for new connections.

How valid is this usecase given that we now send relays to the clients / gateways outside of connections?