Yes
It will work, but as I said may not correspond to the textbook definition(s) of multivariate Laplace. It’s the one defined by that specific generative process. Whether it suits you depends on what properties you want the multivariate generalization to capture.
The feature needed is not laplace specific, it’s for inferring the logp of dot producted distributions (used in the CustomDist) and depends on this PR: Derive logprob of matmul by ricardoV94 · Pull Request #7542 · pymc-devs/pymc · GitHub
It will be available in the next PyMC release, whatever number it turns out to be. I’ll try to make one tomorrow.