design: "Seed" routers and "Bridge" routers

[jamesmunns]

Currently, all router implementations assume they are the One and Only router, which means they all assign net-ids as they feel.

For the near term, we can probably get to the next slate of functionality by requiring the existence of a single Seed Router (to borrow the AppleTalk term), which is responsible for the allocation of net ids.

This probably requires defining well_known endpoints/topics for net id negotiation.

I could see:

• topic broadcast: "who is the seed router? tell X"
• topic broadcast: "X is the seed router"
• endpoint request: "to SEED, BRIDGE wants a net-id"
• endpoint response: "to BRIDGE, SEED is granted NET_ID for DURATION/request rejected"
• endpoint request: "to SEED, BRIDGE wants to retain NET_ID"
• endpoint response: "to BRIDGE, NET ID extended for DURATION/request rejected"

[jamesmunns]

Work is ongoing for this:

• Consider source interface when routing #125 added the ability to consider the "source" interface when routing, to avoid routing packets back to the interface they came in on
• Separate "net id" and "destination route" conceptually #126 started to rework DirectRouter to split "routes" from "interfaces", so we can have indirect routes
• Teach DirectRouter how to act as a Seed Router #129 added initial seed router concepts, allowing for assigning non-local net-ids
• Add socket discovery, implement first seed router services #130 added the "net id negotiation" described above, as well as socket discovery

Currently there are no "bridge routers" yet, though the ergot-bridge-client currently pretends to have another interface to test the net id requesting logic.