Add RFile block info to validate the block was read correctly

Ideas for this:

* Add magic number at the beginning of each RFile block, to account for empty blocks
* Add hash at the end of each RFile block to ensure the block is fully read and doesn't return early, and to ensure there aren't any errors in the block that was just read (may particularly benefit with block caching, which forces blocks to be fully read)

Start by locating the RFile block read and write entry points, including the block-caching path mentioned in the issue. Determine how block boundaries and empty blocks are represented before assessing magic-number and end-of-block hash validation. Done means blocks can be checked for complete reads and detected errors without returning early.

Written by the indexing model from the issue text.