btrfs: make buffered write to copy one page a time

Currently the btrfs_buffered_write() is preparing multiple page a time,
allowing a better performance.

But the current trend is to support larger folio as an optimization,
instead of implementing own multi-page optimization.

This is inspired by generic_perform_write(), which is copying one folio
a time.

Such change will prepare us to migrate to implement the write_begin()
and write_end() callbacks, and make every involved function a little
easier.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>