allow building a kernel without buffer_heads v3

[Christoph Hellwig <hch@lst.de>]

Hi all,

This series allows to build a kernel without buffer_heads, which I
think is useful to show where the dependencies are, and maybe also
for some very much limited environments, where people just needs
xfs and/or btrfs and some of the read-only block based file systems.

It first switches buffered writes (but not writeback) for block devices
to use iomap unconditionally, but still using buffer_heads, and then
adds a CONFIG_BUFFER_HEAD selected by all file systems that need it
(which is most block based file systems), makes the buffer_head support
in iomap optional, and adds an alternative implementation of the block
device address_operations using iomap.  This latter implementation
will also be useful to support block size > PAGE_SIZE for block device
nodes as buffer_heads won't work very well for that.

Note that for now the md software raid drivers is also disabled as it has
some (rather questionable) buffer_head usage in the unconditionally built
bitmap code.  I have a series pending to make the bitmap code conditional
and deprecated it, but it hasn't been merged yet.

This series is against Jens' for-6.6/block branch.

Changes since v2:
 - fix handling of a negative return value from blkdev_direct_IO
 - drop a WARN_ON that can happen when resizing block devices
 - define away IOMAP_F_BUFFER_HEAD to keep the intrusions to the
   iomap code minimal (even if that's not quite my preferred style)

Changes since v1:
 - drop the already merged prep patches
 - depend on FS_IOMAP not IOMAP
 - pick a better new name for block_page_mkwrite_return