Improve how we handle two different set of CARs being uploaded for the same root CID

[dchoi27]

Context (from @vasco-santos):

Reading from freeway after second upload will mean 3 CAR CIDs to go through via freeway.
Freeway does a R2 list with prefix in DUDEWHERE ${rootCid}/ and gets all CAR files ever associated with given Root
We will read all indexes from Satnav for given CARs (1 extra…) and create an in memory Map index, where we keep track of all index files
Then we start the actual read from root. We read from root block and start unixfs export by going through all links from root block we get until full dag
TLDR:
In case there are more side indexes than needed, we will read all the extra side indexes and they will count towards our sub requests limit (and max shards protection)
When actually reading, we try to find a given block CID in multiple indexes in parallel and once we get one CAR CID where CID is we return back…. So, in practise we won’t do extra reads by having more CARs associated with Root CID