Add `notebooks/wang_mle_tutorial.ipynb` walking through pooled + hierarchical estimation

[YichengYang-Ethan]

Problem

The Wang Transform MLE estimator (oracle3.pricing.wang_mle.WangMLE) is the academic centerpiece of Oracle3 (v1.1.0+, DOI 10.5281/zenodo.20062549). The docstrings are thorough, but there is no end-to-end notebook that a new user can step through to see what the estimator actually does, what the output means, and how it connects to trading decisions.

We need a notebooks/ folder with one polished tutorial.

Acceptance criteria

1. Create notebooks/wang_mle_tutorial.ipynb with the following sections:
   1. Synthetic data generation: simulate 5,000 contracts under the model p_mkt = Phi(Phi^{-1}(p*) + lambda) with a known true lambda (e.g. 0.18).
   2. Pooled MLE: fit WangMLE().fit(prices, outcomes), recover lambda_hat, and verify it's within ~2 SE of the true value.
   3. Hierarchical MLE: simulate covariates (volume, duration, extremity) with a known beta vector, run build_design_matrix + fit, print summary_table().
   4. Likelihood-ratio test: compare nested models with WangMLE().lr_test(...).
   5. Greeks visualization: plot kelly_fraction and delta_lambda vs. p_market for a fitted lambda using oracle3.pricing.greeks.
2. Add a one-line entry to README.md under "Examples" linking to the notebook.
3. Notebook should run end-to-end with no errors using only numpy, scipy, matplotlib (already deps), and the package itself. Cell outputs committed.

Hints

• API: oracle3.pricing.wang_mle.WangMLE, MLEResult.
• Empirical priors at the top of wang_mle.py (LAMBDA_POLYMARKET = 0.166 etc.) are useful as anchors when discussing the lambda magnitude.
• Greeks helpers in oracle3.pricing.greeks: kelly_fraction, delta_lambda, compute_greeks_table.
• For sample data generation, scipy.stats.norm.ppf / cdf mirror the model's probit form.

Estimated time

3-4 hours.