Feature Request: Batched convolve2d

Description

Hello, I'd like to request the addition of a batched convolve2d function - one that takes two lists of arrays (or perhaps implemented as two 3D arrays) and convolves them elementwise.

Currently, convolve2d can only handle a single pair of 2D arrays. This leaves much parallelism on the table when multiple independent pairs of arrays must be convolved, as each convolution would be executed in serial, rather than parallel. This is particularly impactful for GPU acceleration, as manually issuing multiple convolve2d calls would result in many additional synchronization points, significantly reducing performance compared to if the convolutions were executed in parallel.

Additional Information

No response