Parallel evaluation of the integrand

In the previous Cubature.jl package, it is possible to evaluate the integrand in parallel by requiring the function `f` to accept matrix arguments. Is it possible such feature can be supported in HCubature.jl?

Start by reviewing HCubature.jl's current integrand API and the prior Cubature.jl behavior for matrix arguments. Define how function f should accept batches and what parallel evaluation means here; done should be demonstrated by coverage for matrix-valued calls and parallel integrand evaluation.

Written by the indexing model from the issue text.