More efficient block reconstruction in withholding attacks via chunk bitmap

Right now, if there's a withholding attack, reconstruction occurs by full nodes "bruteforcing" light nodes with different chunk requests via bitswap, which is very slow for larger blocks, as for a 256x256 square, this would be 100,000+ chunks.

We can make the protocol more efficient by having full nodes ask light nodes for a bitmap of the chunks they have.

Full nodes would send light nodes a request `QueryKnownChunks` to a light node. The light node would respond with a bitmap where each bit would represent a different chunk, and if that bit is set to `1`, then the light node has that chunk. The full node would then be able to query the light node for the specific chunks it has, rather than bruteforcing them.