Peer Discovery For Graphsync Metadata Request

[hannahhoward]

What

To kick off a lightning storm request, we need to make a Graphsync metadata request(s) to a peer who has the content. But even before that, we need to discover which peer has the content.

There are a couple ways we could do this:

• Utilize Bitswap sessions to broadcast the initial want-haves, and expose the sessions ([ipfs/go-bitswap] Exposing the session peers boxo#75)
• Just start with a metadata graphsync request to all local peers (not clear how much duplicate data this actually is)
• Just start with a metedata graphsync request for the root block with a pure matcher selector to limit duplication (not actually a great one-block selector as large bytes nodes could get auto traversed)
• Build an actually graphsync want-have protocol, or peer discovery protocol

The approaches merits further discussion and this ticket should be done after Lightning-Storm has been proven out in testbeds