Moment matching with iwmm package

Currently loo_moment_match only works on stanfit objects (due to the reliance on the log_prob methods from rstan), but not CmdStanFit or just a matrix of draws (see e.g. #209). I was thinking about the best way to expand support of loo_moment_match to other objects, and am opening this issue mostly for discussion at this stage.

@topipa has created a generic implementation of moment matching (iwmm) which works on a matrix or stanfit object. Recently we have also added CmdStanFit support to this package (using the new model methods in cmdstanr which are currently not yet in a release).

iwmm is not (yet) on CRAN, so now might be a good time to discuss the best way to interoperate with loo.
After some discussion with @avehtari, I currently see three options (there may be others I haven't thought of):

1. update functions in loo, no change to dependencies
   Code from iwmm is copied into loo and adapted for use in loo::loo_moment_match (made specific for leave-one-out importance posteriors). iwmm would remain an independent package for generic importance sampling.
2. add iwmm dependency in loo
   iwmm is submitted to CRAN and loo::loo_moment_match is rewritten to use iwmm::moment_match. loo would then import and depend on iwmm.
3. move iwmm functions into loo
   All iwmm functions are moved into loo, and kept generic (i.e. not specific for leave-one-out posteriors). Keeping the functions generic has some precedence as loo is the home of loo::psis which is used in cases other than leave-one-out CV (e.g. in priorsense and adjustr).

@jgabry @avehtari @topipa @paul-buerkner , do you have any thoughts on this?

Start by reviewing loo::loo_moment_match and the generic implementation in the linked iwmm package, including its matrix, stanfit, and CmdStanFit support. Compare the three interoperability options in the issue and determine which design should be adopted; done means the project has a decided approach for expanding supported inputs.