Bring dactyl/matrix.c in line with quantum/matrix.c (#2613)

Bring dactyl/matrix.c in line with quantum/matrix.c (#2613)

* Use the new debounce algorithm in dactyl/matrix.c [#2065]

This incorporates the fixed/optimized debounce code added to
quantum/matrix.c in:

* 508eddf8ba8548d3f71e1c09a404839beb49f45c
* 4c6960835c0a6e29670dabdc27117d7d3c7f99f5
* 32f88c07173b795c6981c779057dceba00aeb1cb
* f4030289744fc6dc82dd85c955070c0845813cc5
* a06115df19a74d39b08758472b221e630c3680d3

* Fix the row/column swap in dactyl [#2065]

With a column-driven keyboard, reading from the mcp23081 returns a
column-state, which takes some extra work to translate into the
row-state used in the actual matrix. The ergodox_ez code sidestepped
that problem by calling rows "columns" and columns "rows." With this
change, the dactyl now calls rows "rows" and columns "columns."

* Cleanup: variable names, documentation [#2065]

* Support MATRIX_MASKED in dactyl/matrix.c [#2065]

* Only unselect one col in unselect_col [#2065]

Bonus: saves one i2c transaction per matrix_scan!

* Implement COL2ROW in dactyl/matrix.c [#2065]

* Fix a typo in dactyl/matrix.c

This entirely doesn't matter. The PORT values are set during
init_keyboard and never change. They're repeatedly set to the same
thing. These PORT lines shouldn't even exist, but since they do, they
should at least look right.

* Implement COL_PINS/ROW_PINS for dactyl [#2065]

* Rename "mcp23018" to "expander" [#2065]

I honestly don't know whether/how well this code works with other I/O
expanders, but at least in theory, it should be generic enough to work
with others. Given that, the variable names shouldn't refer to a
specific model of expander.

* Remove matrix_power_up from dactyl/matrix.c [#2065]

It's commented out in quantum/matrix.c, and the dactyl has no power
up/down behavior beyond being unplugged (which goes to matrix_init), so
there's no sense keeping it around.

* Only initialize expander_input_mask once [#2065]

...and rename input_mask to expander_input_mask, since now that it isn't
scoped to init_expander it isn't clear that it's only for the expander.