Shannon and Simpson Diversity

From category counts, implement:

• shannon(counts) = $-\sum p_i\ln p_i$ (natural log; zero counts ignored)
• simpson(counts) = $1 - \sum p_i^2$: the chance two random picks (with replacement) differ
• effective_number(counts) = $e^{H}$ (Hill number of order 1): "as diverse as this many equally common categories"