Editorial: Clarify namespace boundaries

[bumblefudge]

Contributors have been tripped up a lot recently on what a new namespace is in two different ways:

• An L2 with new capabilities, different consensus, etc could be a namespace unto itself, or just a set of caveats/additions to the namespace of its L1-- the major difference seems to be whether the L2 speaks its own RPC (in addition to the L1s). Does this mean there is a 1:1 RPC<>namespace definition that should be made more explicit?
  • corner case: as more and more L1s support "EVM-mode" (a second set of RPC commands and endpoints), and particularly when only ONE chain within that namespace (e.g. Moonbeam in Polkadot) does so, should there be some kind of new CAIP-X for "extensions"? would evm/caipx.md become an infinite list of special cases, if so?
• should the CAIP-104 and/or the template make all of this more explicit?

[bumblefudge]

Addressed by #81. See new FAQ section