ExactComputer Refactor Types and Making n_players optional.

[mmschlk]

Motivation and Context

This PR closes #388 and improves the ExactComputer's code quality by making the index parameters in method signatures more tightly bound by introducing Literals (e.g. replacing index: str with index: Literal["SV", "BV"]). This reduced in a couple of tidy up operations (introducing match statements for reducing the scope of the variables and pyright compliance.

❓ Remaining Questions:

In ExactComputer.__call__ we compute the Moebius (computed with ExactComputer.moebius_transform) and Co-Moebius (computed with ExactComputer.shapley_base_interaction) while both can be computed via ExactComputer.shapley_base_interaction. Should we harmonize this (see TODOs)? If so should we remove the ExactComputer.moebius_transform property? As I see it now, Moebius is always getting computed with all orders but Co-Moebius not (likewise to all other indices available in ExactComputer.shapley_base_interaction.

---

Public API Changes

• No Public API changes
• Yes, Public API changes (Details below)

---

How Has This Been Tested?

Added new tests for the cases where the indices may be misspecified and adapted some tests to adhere to the new method signatures.

---

Checklist

• The changes have been tested locally.
• Documentation has been updated (if the public API or usage changes).
• An entry has been added to CHANGELOG.md (if relevant for users).
• The code follows the project's style guidelines.
• I have considered the impact of these changes on the public API.

---

[FFmgll]

computation of moebius transform through moebius_transform() is faster than using shapley_base_interaction(). The latter uses a loop over powerset(grand_coalition) and within loops over powerset(grand_coalition, max_size=k), where k=n for Moebius whereas moebius_transform() requires the inner loop only over powerset(subset from outer loop). While shapley_base_interaction is convenient to compute different values, I assume that other computations could also be quicker with other loops, e.g. Co-Moebius.

The base_interaction() function also allows to only compute up to order k, maybe we could add this to moebius_transform(). In general, I would support having separate functions for each index that are optimized for each index.

[FFmgll]

rest lgtm