Hey @Webmaster4o, sorry about the late response!
When iOS 8 came out, I tinkered a little bit (quite a lot actually) with custom keyboards, but my experiments never really left the rough prototype phase for a couple of reasons.
My idea was basically to make a "BYOK" ("build your own keyboard") app, separate from Pythonista, in which you could completely customize key layout, and trigger script actions, Editorial-like workflows, or text snippets via special keys or maybe abbreviations (similar to TextExpander)... By default, it would look similar to the standard keyboard, but you'd be able to rearrange all the keys, and change their mapping. I basically wanted the thing to be usable as a full replacement for the system keyboard... Unfortunately, it turns out that replicating the system keyboard's functionality is extremely difficult, in particular getting auto-correction reliable enough for an even halfway decent typing experience. Before you even get to stuff like predictive text, the system keyboard does quite a bit of magic that you don't really notice until it's gone, e.g. invisibly adjusting the touch targets of keys based on statistical language data (e.g. the "U" key invisibly gets larger after you type "Q", but that's just the simplest example – it also uses individual typing speed/accuracy data etc.). The kind of research that goes into touchscreen keyboards is actually quite fascinating, but the point is, the app that I had in mind turned out to be not realistically feasible for me.
Another source of frustration was that the custom keyboard APIs in iOS 8 were incredibly buggy and limited. I'm not quite sure about the bugginess in iOS 9 (I think the situation may have improved a bit), but as far as I know, the limitations are largely still the same, which means that the system keyboard can do things that custom keyboards just can't implement at all. Since the system keyboard gained new capabilities in IOS 9, these limitations are actually even more frustrating. For example, the new trackpad gestures are simply impossible to implement in a custom keyboard because you can only move the cursor left or right, not up or down... The thing I find most annoying is that switching from one keyboard to another takes far too long if you have a couple of keyboards installed, because from a custom keyboard, you can only switch to the next one, whereas on the system keyboard, you can hold the 'globe' key to switch to any keyboard directly (there's still no API to do that in a custom keyboard).
This last point was actually the major reason that I wanted to implement a "full" keyboard, not just a special-purpose one (like the various alternative Emoji/Unicode pickers). I just find it way too annoying to switch between keyboards, and while it's gotten a little better simply because the devices are faster now, I found that I just don't like even having third-party keyboards installed at all because once I switch to one of them accidentally, it takes forever to get back to the standard keyboard (made worse by the fact that third-party keyboards often have their "next keyboard" buttons in different places).
Aaaanyway, if I understood you correctly, the thing you have in mind would actually be closer to a "blank canvas" keyboard that you could program yourself, using the ui module etc. This would require a couple of changes in my Python modules because some system capabilities aren't available in a keyboard (app extension), but some of that is actually the same as for the action extension in the current beta... There are lots of details that I'm not sure about, e.g. what would happen with console output, how would text entry be handled (I guess there would need to be a "keyboard within the keyboard"?) etc... Also, the capabilities of Apple's APIs for custom keyboards (that I'd expose in a "keyboard" module or something like that) are probably more limited than you think. For example, it is not possible at all to get the currently selected text, you can basically only get the previous/next word (sometimes not even that, depending on the cursor position). You also cannot trigger actions like "copy", "cut", or "undo" in the edited document (pasting is possible though, by simply "typing" the text in the clipboard). Cursor movement is limited to left/right, and there isn't a way to know where the cursor currently is, relative to the current line or the whole document (so you couldn't implement things like "jump to the end of the line")...
Given all that, I do wonder how useful a thing like this would actually be. I admittedly have a hard time coming up with concrete use cases that aren't basically gimmicks, but I imagine you have some ideas there?
Wrapping up, I do think that a "blank canvas" custom keyboard, as part of Pythonista, would be feasible – I definitely don't think it's something I'd start working on before 1.6 is out (as you might have guessed), and I'm honestly not sure at this point if it would be useful for enough people to make it actually worth doing.