[numba.md] Update np.random → Generator API

[Chihiro2000GitHub]

Summary

This PR migrates legacy NumPy random API usage in numba.md as part of QuantEcon/meta#299.

Details

The Numba/JIT-related classification and changes in this PR follow the guidance at https://manual.quantecon.org/styleguide/code.html#numpy-random-number-generation. I would be grateful if you could refer to the Numba section of this page when reviewing.

Case A (update(), main text, @jit): Left unchanged. Although update() is @jit-decorated, it is called from a prange loop in compute_long_run_median_parallel, making it unsafe to pass a shared Generator. Flagged for reviewer judgment.

Case B (speed_ex1 solution, @jit / no parallel): rng = np.random.default_rng() placed before the @jit definition; signature changed to calculate_pi(rng, n=...); np.random.uniform → rng.uniform. Call sites updated. Note: passing a Generator into a @jit function may require Numba to use object mode. I checked that the updated code runs as expected on my side, but I would appreciate reviewer confirmation.

Case C (speed_ex2 solution, plain Python / later JIT-compiled via jit(compute_series)): Same pattern as Case B. Signature changed to compute_series(n, rng); np.random.uniform → rng.uniform. Both the pure Python and jit-compiled call sites updated.

Case D (numba_ex3 solution, @jit(parallel=True) + prange): Draws lifted outside the prange loop. rng, u_draws, and v_draws defined before the @jit(parallel=True) definition; signature changed to calculate_pi(u_draws, v_draws); loop length derived from len(u_draws). n = 1_000_000 kept to match the original example size.

Case E (numba_ex4 solution, @jit(parallel=True) + prange): Legacy np.random.randn() calls kept intentionally as a narrow memory-constrained exception. Pre-allocating (M, n) shock arrays with M = 10_000_000 and n = 20 would require approximately 3.2 GB, which is likely beyond what most readers' machines can comfortably accommodate. A short comment has been added inside the loop explaining this.

Hi @mmcky and @HumphreyYang, I'd be grateful if you could take a look when you have time.

[github-actions]

🚀 Deployed on https://6a025d7400b6b900d7d3e8af--epic-agnesi-957267.netlify.app