Cleanup of the VFS block file system

[nfeske]

I think that the VFS block file system is not in a good shape. I would like to see it improved in several respects:

• The _block_io method should be replaced by _read and _write methods. Handling both operations in one method makes things needlessly complicated and inconsistent. I.e., the misleading error message "Could not read block(s)" that could also appear on write operations, or the use boolean write argument, and the optional bulk argument.
• The handling of partial block accesses is hard to comprehend and depends on questionable conditions. In particular, I was wondering about (count % _block_size) >= 0. Isn't this always true?
• It should use the block session asynchronously, once the VFS supports this.
• It should use the blit library instead of byte-wise memcpy.