[Multi-Tenant] ILP Addresses per Tenant

[mkurapov]

@nathan said:

I'm having a bit of trouble dealing with the asymmetry of peering between a multiple tenants on instance of Rafiki and a single tenant on another instance.

Given a Rafiki instance (A) with two tenants (A-1 and A-2), and a tenant on another instance (B-1), consider the scenario where we are trying to establish a peering relationship between the following:

• A-1 <-> B-1
• A-2 <-> B-1

Since each peering relationship needs to establish a corresponding pair of incoming/outgoing http tokens, and meet a uniqueness constraint for those tokens, establishing the peering relationship on the B-side gets a little tricky.

• Creating two peers on (B-1) causes a conflict where both entries point to the same ILP address using the same asset.
• The peers for (A-1) and (A-2) can't use the same incoming token coming from (B-1) because it also creates a conflict.

It might seem apparent to loosen the uniqueness constraint for httpTokens to allow a tenant id to distinguish them further, but I'm pretty sure on the connector level we wouldn't be able to apply this distinction. Thoughts?