Is there appetite for PRs adding metal kernels of scipy.special functions used in probabilistic programming, such as:
Context: I'm a dev of pymc/pytensor, we're interested in expanding MLX support as a compute backend for bayesian models. These functions are important for computing the logp of various distributions.