spi_bfin: wait for tx to complete on some cs_chg paths

PBX 2 SPI devices need the nonstandard "cs change per word" mechanism.

This patch is one of three updating this driver to make the last data bits get
sent before advancing the transfer ...  in this case, before the chipselect
gets deactivated.

Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>