Data-structure to use caching the header of the NEXT block. Only works if next read
that comes in here is next in sequence in this block.
When we read, we read current block and the next blocks' header. We do this so we have
the length of the next block to read if the hfile index is not available (rare).
TODO: Review!! This trick of reading next blocks header is a pain, complicates our
read path and I don't think it needed given it rare we don't have the block index
(it is 'normally' present, gotten from the hfile index). FIX!!!