Following commit d0a22a4b ("gfs2: Fix iomap write page reclaim deadlock"),
gfs2_iomap_begin and gfs2_iomap_begin_write can be further cleaned up and the
split between those two functions can be improved.

With suggestions from Christoph Hellwig.
Signed-off-by: NAndreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: NBob Peterson <rpeterso@redhat.com>