Because of errata i689 the keyboard can idle with state where no key
up interrupts are seen until after the next key press.

This means we need to first check for any lost key up events before
scanning for new down events.

For example, rapidly pressing shift-shift-j can sometimes produce a J
instead of j. Let's fix the issue by scanning the keyboard in two
phases. First we scan for any key up events that we may have missed,
and then we scan for key down events.

Let's also simplify things with for_each_set_bit() as suggested by
Dmitry Torokhov <dmitry.torokhov@gmail.com>.
Signed-off-by: Tony Lindgren <tony@atomide.com>
Link: https://lore.kernel.org/r/20210110190529.46135-3-tony@atomide.comSigned-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>