Changes to monitor proto

[rvolosatovs]

I propose 3 changes to monitor proto:

1. Let's rename GatewayUplink and GatewayDownlink to RouterUplink/RouterDownlink respectively, to better reflect the source of data and type.
2. In current implementation ttn opens 3 streams per component(gateway,router,handler,broker) at all times. Let's open only 1 stream per component and wrap message in a oneof Status,Uplink or Downlink. This makes handling much easier.
3. Let's add all missing endpoints to the components, in other words, let's make sure there are 3 endpoints per each component(or 1, if we agree on my previous point)

[htdvisser]

1. Renaming this will break existing clients, so that's not going to happen (gRPC service endpoints are addressed by name).
2. That shouldn't make much of a difference, but we could add a "composite" stream in the future. Not really a priority imo.
3. Not all endpoints make sense. For example, the NetworkServer component is mostly just a database and doesn't really add additional information compared to the Broker

Considering that we're also thinking about merging some components (router,broker,networkserver) I don't see why we would make these changes.

[rvolosatovs]

1. This is 2 calls to gorename in ttn repo, 1 call to gorename in NOC repo. Which clients are going to be broken? TTN will just get an error about an unimplemented endpoint for a couple of seconds.
2. Given the functionality we want, it does. It's much easier to count uptime that way, if we rely on stream status, also why would gateway authenticate 3 times if 1 time is enough?
3. ok

[htdvisser]

Please keep in mind that are APIs are considered stable for the v2 version. We can't just go around and change things because we realize that it wasn't perfect.

[rvolosatovs]

can we do this for v3 then?

[htdvisser]

Maybe. Iceboxing this issue