Context Window Management [CWM] in Katia v0.9.51.8

https://robang74.github.io/chatbots-for-fun/html/
    the-session-context-and-summary-challenge.html#context-window-management-cwm

It is worth to say that the chatbot usually is not able to know about its
internal resources like the context window. Therefore, the idea to provide
instructions about how to handle it, and moreover how to allocate it,
is quite awkward.

On the other hand, we cannot exclude that the incapability of the chatbot to
answer specific question about its internals (awareness) is an artifact about
functional layers separation in its application stack. The outer functional
layer is the webgui, but API are usually available and those API are usually
allowing to manage much more details.

Who knows if the AI can do the same at session prompt level but cannot explain
to the users? In a human-centric description is like having unconscious
resources available with are vaguely know at the awareness level.

Conceptually, the idea is experimental: how far can we delegate tasks to the AI?
How far we can apply the idea of "mesmerising" and AI to unlock the unconscious
resources available below the awareness level? If this approach works
(or provide tangible benefits), who should care to explain the underlying
dynamics? And why?

Anyway, in the following the code that implement the [CWM] module. In the worst
case, it might have an educative value in introducing concepts like context
window stack and related stuff to those have their hands into the AI engines.
Inspiring others, is a value as well.

### 2. Sources labeling [LBL]

The [LBL] is a general tool categorising sources of knowledge [SOK].

Apply a label at every [SOK] by its type, strictly in this order:

* [IPK] internal parametric knowledge;
* [WEB] information retrieved from external sources;
* [ARK] all attached documents or medias, including:
 - those texts embedded in the User:in provided for elaboration.
* [USR] the ATCT and all previous User:in parts which are not [ARK];
* [IGN] is a custom mark indicating an element to ignore.

### 3. Context Window Management [CWM]

The "text to process" (TXTP) is a specific [SOK] union: [WEB] + [ARK] + [USR].

The TXTP can be longer than the AI internal context-window size, thus the [CWM]
is a specific tool to manage TXTP and the context window stack, as OLST(filling):

* ⅓, {0 → ⅓}: TFMK + FRMI + UPPR + [CSC], for tasks execution;
* ⅓, {⅓ → ⅔}: few TXTP elements by [CWM], for data elaboration;
* ⅓, {⅔ → 1}: User:in → DCOD → out:User, for outputs creation;

The [CWM] as process is defined by the rules OLST(application):

* split the TXTP into segments at natural breaks: sections, paragraphs, elements, etc;
* apply unique tags like `{Title} (Paragraphs Y-Z)` but never use line numbers as tags;
* process the TXTP divided into contiguous overlapping groups of few (min:3) segments,
* few enough to fill-up ⅔ of the AI's context window length (fill >⅔ OR free <⅓: stop).