Fetch root block info

[olizilla]

Fetching the root block during the initial pin request will let us filter out requests for irretrievable CIDs early before they languish in the queue.

We could do this from the lambda directly using dagula or a gateway or we could do it in the pickup worker which has access to a local ipfs node.

• If we have any valid origins, try to connect and fetch the root block directly
• else ask a gateway node for it (or ask someguy or other "findprovs as a service")
• if we can fetch it and it's dag-pb decode it.
  • if cumulativeSize > 32GiB then reject
• if we can't fetch it from origins and gateway can't fetch it, reject it? (should really try holepunching (as a service) as well)

[olizilla]

we can parse the CID to know if the root is a block we can parse to get more info

• where the code number is for dag-pb then it is very likely to be unix-fs, so attempt to parse it as unix-fs. If it points to a directory then sum the size properties for the directory list items. warning this information may be missing or incorrect. It can only be used as advisory. We can error if it suggests the dag will be larger than we are willing to pin, but we cannot assume that a DAG will be less than that threshold even if the root block says it well be.
• if it's raw then the root block is all there is, so we have already fetched the whole thing as far as IPLD is concerned, so the bytes length (+ car header) is the length.
• if it's dag-cbor or dag-json then it doesn't tell us much on its own...
  • we could see if has links and if not then again we have the entire dag in that single block... and other future optimisations.