libmfu: optimize sparse writes during mfu_flist_copy

[adammoody]

When executing an mfu_flist_copy to copy files with the sparse option enabled, we currently detect blocks of data that are all zero and skip writing those portions. If we're using large blocks, we could look to optimize this further.

For example, what if we're using 64MB size blocks and only the first byte in the 64MB is non-zero. We could just write that one byte and avoid writing the remaining (64MB - 1) bytes.