distributional 0.8.0

New features

• dist_mixture() gains a type argument supporting both probability mixtures
  (type = "probability", the default, existing behaviour) and quantile
  mixtures (type = "quantile"). A quantile mixture (Vincentization) defines
  the quantile function as a weighted average of the component quantile
  functions: Q(p) = sum(w_i * Q_i(p)). Its CDF and density are computed
  numerically from this quantile function.

• Added dist_convolved() for the distribution of the sum of two independent
  random variables, computed via FFT-based numerical convolution (#106).
  The + and - operators now automatically produce a dist_convolved object
  for distribution families without a known closed-form sum. Distributions with
  known closed-form sums (e.g. dist_normal(1, 2) + dist_normal(3, 1)) continue
  to produce exact results.

• Added dist_quantile() as the primary interface for the quantile/percentile
  distribution, accepting probabilities on the [0, 1] scale.
  dist_percentile() is now an alias that accepts the [0, 100] scale (#116).