parrallelize posterior particle loading

[arik-shurygin]

as requested in #311 figure out a way to parallelize the load_posterior_particle() function which calls run_simulation. Either with Pool() or with jax.pmap().

The main issues faced when trying this earlier (and why it is not included in #316 ) is that Pool() requires pickle serializibility and pmap only works with jax operations, which our code breaks both of in its current architecture. Maybe I am missing something here.