feat: add categorical distribution

[kholdrex]

Summary

Adds a categorical distribution for finite discrete choices, expanding the existing discrete distribution API beyond Bernoulli, Binomial, and Poisson. The implementation accepts non-negative category weights, normalizes them, and samples category indices with deterministic seeded RNG support.

Changes

• Add Categorical with parameter validation, normalized probability access, PMF/log-PMF, mean/variance, and sampling through DiscreteDistribution.
• Cover analytical probabilities, zero-probability categories, invalid weights, seeded sampling, and trailing zero-weight sampling behavior.
• Update the discrete distribution example and README to show categorical usage.

Test plan

• Rust format check passed.
• Strict Clippy passed across all targets and features.
• Full test suite passed with all features.
• Rust documentation built with warnings denied.
• Benchmarks compiled successfully.
• Independent review found no blocking findings.