Scalib

-------------------
* Add ``scalib.preprocessing.Quantizer`` for conversions of floating-point
traces to int.
* Support modular subtractions (generalized summation) in FactorGraph.
* Raise minimum supported python version to 3.9.

-------------------

* ``scalib.attacks.FactorGraph``:

* Add ``GENERIC`` factors

* Add clear beliefs flag for BP.

* Remove divisions in belief propagation (improves numerical robustness, fixes some NaN issues).

* Support python 3.12 in releases.

-------------------

* Add ``scalib.modeling.RLDAClassifier`` and ``scalib.metrics.RLDAInformationEstimator``.
* Fix build dependences and run build in an isolated environment.

-------------------

* Fix ``Distribution`` size bug in belief propagation computation (102).
* Do not include debug symbols in release builds (makes wheel much smaller).
* ``Ttest``: do not crash on non-contiguous/fortran-order arrays
* Improve examples and README

-------------------

* Run CI on Mac Os and build wheels (x86_64 and arm64, but we test only x86_64). No AVX2 due to old runner in github CI.
* Add ``scalib.tools.ContextExecutor``, as a solution to ``LookupError`` in
``get_config()``.
* Fix numerical underflow in ``BPState`` when multiple traces are used.
* Fix missing import in ``MultiLDA``.
* Run ``BPState`` methods on the threadpool.
* Make threadpool initalization lazy -- makes SCALib play more nicely with ``ProcessPoolExecutor``.

-------------------

* Add ``vars`` and ``factors`` methods on `FactorGraph`.
* Add ``fg`` property on `BPState`.
* Fix bug in belief propagation. Numerical instabilities could lead to negative
probabilities for AND (and probably XOR and ADD) factors.