[ENH] Add probabilistic regressor with shrinking intervals

[arnavk23]

Reference Issues/PRs

Towards #7

What does this implement/fix? Explain your changes.

This PR adds a new probabilistic baseline estimator ShrinkingNormalIntervalRegressor for interval and quantile prediction.

• In mean_sd mode, intervals are built as mean ± z * sd / sqrt(n), where z is the Normal critical value for the requested coverage.
• Because standard error scales as 1/sqrt(n), interval width decreases as sample size grows, giving a simple demonstration of shrinking uncertainty with more data.

What is implemented.

• A feature-agnostic probabilistic regressor that estimates target location/spread from training targets.
• Predictive intervals via predict_interval and predictive quantiles via predict_quantiles.
• Two modes:

1. mean_sd: Normal-approximation intervals/quantiles with shrinkage in n.
2. quantile: empirical quantile baseline from training targets.

Limitations

• mean_sd assumes an approximately Normal target distribution around a global mean and uses a single global spread estimate.
• quantile mode is static. It does not shrink with n by construction.
• The estimator ignores feature effects in uncertainty (and in location beyond a global mean) so it is a calibration/baseline-style method rather than a full conditional model.

Does your contribution introduce a new dependency? If yes, which one?

No new dependencies.

What should a reviewer concentrate their feedback on?

1. Correctness of the shrinkage logic in mean_sd mode, especially width scaling with 1/sqrt(n).
2. Clarity and appropriateness of the two-mode API (shrinking Normal mode vs static quantile baseline).
3. Whether the documented limitations are sufficiently explicit for users.

Did you add any tests for the change?

Yes.

• Added focused tests for interval shrinkage with increasing n in mean_sd mode.
• Added tests for predict_interval and predict_quantiles shape and value correctness.
• Added a small-n edge-case test to ensure stable finite outputs.
• Generic estimator suite coverage remains in place.

Any other comments?

This estimator is intentionally simple and educational as a probabilistic baseline. Feedback on API naming, mode semantics, and whether to keep both modes in one class versus split classes is especially welcome.

PR checklist

For all contributions

• I've added myself to the list of contributors with any new badges I've earned :-)
• The PR title starts with either [ENH], [MNT], [DOC], or [BUG].

For new estimators

• I've added the estimator to the API reference - in docs/source/api_reference/regression.rst, follow the pattern.
• I've added one or more illustrative usage examples to the docstring, in a pydocstyle compliant Examples section.
• If the estimator relies on a soft dependency, I've set the python_dependencies tag and ensured dependency isolation.