Convergence diagnostic - Rhat is computed by "Rank" method which is different from the method mentioned in documentation?

The Rhat statistic is computed using ArviZ rhat() which can be seen here .
So, as shown in the ArviZ code, the default method is “Rank” which is suggest by Aki Vehtari (2019).
That means the method, pm.gelman_rubin mentioned in the documentation is not being used to compute the Rhat right? Since it is referred to gelman-rubin (1992) and brooks-gelman (1998).
Another question:
How can we change the method to compute the Rhat ? For example, I want to use “split” instead of default method “rank”.