Clarify README "How lorepy works" section

[sepro]

Motivation

• Make the README mathematically precise and correct the misleading statement about confounders so documentation matches the code (default classifier behavior, prediction grid, and how confounders are handled).

Description

• Update README.md "How lorepy works" to state lorepy fits scikit-learn's default LogisticRegression() (L2, C=1.0) unless a custom classifier is provided, that class probabilities are computed via the softmax formula and evaluated at 200 evenly spaced x points, that confounders are included during fitting and then fixed to user-specified reference values at prediction time (i.e., curves are conditional on those reference values), and to clarify how sample dots are generated.

Testing

• Attempted to run python -m pytest -q, but the run aborted due to a missing dependency (ModuleNotFoundError: No module named 'numpy'), so automated tests could not be executed in this environment.

---

Codex Task