btrfs: don't balance btree inode pages from buffered write path

btrfs: don't balance btree inode pages from buffered write path

The call to btrfs_btree_balance_dirty has been there since the early
days of BTRFS, when the btree was directly modified from the write path,
hence dirtied btree inode pages. With the implementation of b888db2bd7b6
("Btrfs: Add delayed allocation to the extent based page tree code")
13 years ago the btree is no longer modified from the write path, hence
there is no point in calling this function. Just remove it.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>