Fixed bug in do_code16

[ofples]

The do_code16 function would falsely send both right and left modifiers for a given keycode because of the way the keycode is checked against the modifiers (bitwise and).
See #865 for an example where this matters.

[fredizzimo]

Yes, this seems like a valid fix to me. But only because the current design doesn't allow programmatically sending a left hand modifier and right hand modifier at the same time in other cases either, as I reported in #524 for example.

The best fix would be to add one more bit to the modifiers and change the do_code16 to handle all three cases. But that would probably require a lot of other fixes too, so that's definitely not needed to fix this bug. There would easily be room for that in the quantum_keycodes enum, by for example allowing "only" 2047 functions and macros instead of 4095.

So the relevant part of quantum_keycodes would look like this.

QK_MODS               = 0x0100,
QK_LCTL               = 0x0100,
QK_LSFT               = 0x0200,
QK_LALT               = 0x0400,
QK_LGUI               = 0x0800,
QK_RCTL               = 0x1100,
QK_RSFT               = 0x1200,
QK_RALT               = 0x1400,
QK_RGUI               = 0x2800,
QK_RCTL               = 0x2100,
QK_RSFT               = 0x2200,
QK_RALT               = 0x2400,
QK_RGUI               = 0x2800,
QK_MODS_MAX           = 0x2FFF,
QK_FUNCTION           = 0x3000,
QK_FUNCTION_MAX       = 0x37FF,
QK_MACRO              = 0x3800,
QK_MACRO_MAX          = 0x3FFF,

[ofples]

I agree that this would be the more thorough approach.
I'd have to check what other side effects the change would have.
I consider this pull request more of a temporary fix, so that tap dance would work properly.

[jackhumbert]

Cool - I'll go ahead and merge this then. Thanks!