Bugfix - ensure keymaps dictionary exists before using it

[mnussbaum]

I realized I might have introduced a bug in #851, I haven't been able to trigger it but I wanted to be safe rather then sorry.

The KeyMap.Add and KeyMaps.Remove functions can't use KeyMap._all to fetch the key map dictionary since they need to modify it, and it appears the dictionary returned by the KeyMap._all function is a copy, and modifying the copy doesn't modify the underlying s:keyMaps variable. However, using the s:keyMaps variable directly is error prone since the variable will be undefined until the very first invocation of KeyMap._all.

This PR invokes and discards the result of KeyMap._all prior to adding or removing any key maps from s:keyMapsdirectly, just to ensure s:keyMaps is defined.

[lifecrisis]

Rather than add more boilerplate to ensure that s:keyMaps exists, why not just define it at the top of the script? It has to be a script-local variable anyway, so we might as well define it where everyone can see it immediately.

Note that this would allow us to dispense with any checks for the existence of the s:keyMaps variable.

[PhilRunninger]

Then we wouldn't need the s:KeyMap._all() function either. Just use the s:keyMaps dictionary instead.

[mnussbaum]

Good call, I've amended the commit to just initialize the s:keyMaps dictionary at the top of the file and totally removed the KeyMap._all function