Weighted samples

I couldn't find any information in the docs regarding how weighted samples (e.g. from SMC or importance sampling) are handled. Should weights be stored as part of the [`sample_stats`](https://python.arviz.org/en/stable/schema/schema.html#sample-stats)?

Start with the linked ArviZ schema documentation and review how sample_stats is defined for stored inference results. Determine whether weighted samples need a schema change or only documentation, then document the agreed representation and behavior for SMC or importance-sampling weights.

Written by the indexing model from the issue text.