Sorter feeds data into the checksum calculator in two different ways: for v1 each specific range of bytes consumed by deserialization is fed in, and for v2 each block read from disk is fed in. This was possibly a minor optimization for limited/top-k sorters which would not always consume the entire block, but v1 checksums are now only used for resumable index builds which always consume the entire input. As such this code can now be simplified to always work in the v2 manner (which produces the same checksum values as the old behavior for v1).

This also enables moving checking the checksum to the file read step rather than closeSource(), which allows removing closeSource() entirely (and openSource(), which has been unused for a while).