Add horseshoe-prior MCMC demo notebook for Bayesian feature selection

[Copilot]

• Explore repository structure, APIs, and existing docs/tutorials
• Run baseline lint/tests to identify pre-existing issues
• Create a new demo folder with a comprehensive MCMC-focused horseshoe prior notebook
• Include markdown explanations of Bayesian GLM + regularized horseshoe math with equations and intuition for sparsity
• Add runnable simulated-data workflow showing model fit, feature selection, GPU/CPU setup, and ArviZ diagnostics/visualizations
• Add a hands-on exercises section with creative challenge prompts highlighting strengths/limitations of the method
• Execute the notebook end-to-end and fix any issues that arise
• Re-run relevant checks for changed files and summarize any remaining unrelated baseline failures
• Run parallel validation and address any actionable findings
• Create a pull request for review

[Copilot]

Pull request overview

Adds a new demo notebook that walks through Bayesian feature selection using a regularized horseshoe prior with MCMC, aligned with the library’s HorseshoeGLM + InferenceConfig APIs and using ArviZ for diagnostics/visualization.

Changes:

• Introduces an end-to-end simulated-data MCMC workflow using HorseshoeGLM (fit, feature importance, recovery metrics).
• Adds ArviZ-based convergence diagnostics and posterior visualization for global parameters and selected coefficients.
• Includes an optional real-data (breast cancer) template plus exercises and takeaways.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.