Return correct file sizes from xFileSize about sqlite-s3vfs HOT 1 CLOSED

After going round the houses far too much, I finally realised that SQLite doesn't leap ahead and write pages without having written previous pages first. So in terms of blocks, never padding any block with zero bytes means that all blocks will end up the same size except the last one.

So now (as of v0.0.10):

• xFileSize is accurate, and is done just by summing sizes of objects in S3
• serialization is similarly unaware of anything page or block size-related, and is done just by concatenating the objects
• and truncation (which is called by VACUUM) works as expected, without any block or page-size knowledge

from sqlite-s3vfs.