optimize: files writes should be block-sized

[steveyen]

On both the regular persistence pathway and the compaction pathway, moss performs writes block-aligned with the starting byte of the to-be-written buffer. But, what about the end of the buffer?
Need to check/review the codepaths to see if moss is writing out block-sized buffers. If not, the filesystem/os might need to perform a wasteful read-modify-write of the last block.