Optimisation

[sbfnk]

Stan supports optimisation for point estimates. This might be a nice feature to add (with appropriate caveats returned to the user).

[sbfnk]

A better option might be to use the laplace approximation within the cmdstanr backend. This will add at least approximate uncertainty.

[seabbs]

or wait a bit and just support pathfinder

[sbfnk]

or not wait as we can use it in the cmdstanr backend? (just finishing local tests, PR forthcoming)