Sn32 rgb v9000. Now with software PWM and less artifacts (qmk#312)

* sn32: revamp the shared RGB driver

* split the rgb logic state config for NPN/PNP
on the non-PWM side of the rgb matrix

* update ROW2COL, cleanup high/low logic

* bump the counter after the scan

* remove atomic blocks
these are not needed for SN32

* formatting

* only reset the counter on wraparound

* fix stray bracket

* ROW2COL: offset cols before the RGB scan

* software pwm control

* fix row2col b/g swap

* make it build

* some pwm updates

* simplify the soft pwm

* small cleanup

* software pwm: move the gpio status change in the handler

* more driver updates

* fix out of bounds bug

* fix one side of the indexing

* sw pwm: update the handlers

* software rgb: full color output

* row2col index fixup: attempt qmk#1

* Revert "row2col index fixup: attempt qmk#1"
don't drink and code kids
This reverts commit b61bf61.

* fixup led indexes

* circumvent stupid hardware decisions
thanks evision

* sw pwm: adjust period based on brightness

* minor fixup

* sw row2col: keep led rows as output

* Revert "sw pwm: adjust period based on brightness"
not having a period big enough for the whole matrix pass is not a good idea.
cols/rows end up not getting scanned and stay dark

This reverts commit e2e2a03.

* actually disable the b/g channels

* only check for OOB access in software COL2ROW

this fixes indexing issues and things are now inline

* prettify and lint it