Erasure coding

As mentioned in #501 , we need to enable data fetch between the compute committees in two consecutive epochs. What if we have some packet loss, or some of the nodes to fetch data from go down.

The easiest way to fix this fault-tolerance issue is by using Forward Error Correction (FEC) coding scheme in wireless communication domain. We can directly use existing FEC coding schemes like erasure coding, spinal codes, etc.

The basic idea of FEC coding is to add some global redundancy on each data packet. Given the condition we may have some packets loss or data sender goes down, if we can successfully receive any N packets from any nodes, we can still recover the whole dataset (say with size of K packets), where N > K but just 1.1-1.3x the size of K.

Start by reading the requirements referenced in issue #501, then clarify how compute committees exchange data across consecutive epochs and how packet loss or unavailable nodes should be handled. Done should mean that the dataset can be recovered from any N received packets, with N only modestly larger than K, but no implementation files or tests are identified here.