This adds a new "KeyFeed" monad which is reminiscent of a parsec-type
monad. This allows keys like 'g' to be mapped using a subbind and the
actual WML part be handled in the catch-all handler.

This also significantly cleans up the typing and complexity of the Wml
implementation.

Create new Grab and KeyCodeMapping modules to support keycode mapping.
That subsystem did not belong in Dsl2.

The code is a bit of a mess, and should probably be moved out of Dsl2
and into a dedicated place, but it works.

I had to do a bit of a hack to get around XMonad's ungrabbing the
keyboard after a Mapping event, which is not the best, but I don't have
a better way of doing it.

Binding to keycodes is good for nonmnemonic key bindings -- where the
choice of key is due to its position on the keyboard rather than the
character associated with it.

Right now only window bindings and subbindings can use keycode bindings.
Root bindings can still only be keysyms and buttons.

I've been using this feature to map some movement keys to Hyper. This
emulates the function key on my M770 keyboard where fn+ijkl are used as
arrow keys. I use the tab key as my hyper key. With xcape it can operate
as a Tab key when release, or a modifier key when held down, which is
awesome.

This is more general than it was before. It's quicker than typing its
synonymous equivalent, <M-f>,. as a bonus it's the same on both dvorak
and qwerty keyboards.

This function invokes a handler if a WML workspace is entered, or if a
non-Wml key is entered, it invokes a different handler.

This allows Wml-tied keys like 'g' to handle non-wml sequences. I.e. "g
<F1>" now displays help.

This new DSL is cleaner and more powerful. This new DSL allows mixing
key and mouse bindings in submappings, which can be very useful.