Rank behaves differently from other systems

Something interesting about the rank function in arrow-rs is that it seems to behave differently compared to others.

Postgres:

> rank () → bigint
> Returns the rank of the current row, with gaps; that is, the row_number of the first row in its peer group.

- https://www.postgresql.org/docs/current/functions-window.html

DuckDB:

> Description The rank of the current row with gaps; same as row_number of its first peer.

- https://duckdb.org/docs/sql/functions/window_functions.html#rank

SQL Server:

> If two or more rows tie for a rank, each tied row receives the same rank. For example, if the two top salespeople have the same SalesYTD value, they are both ranked one.

- https://learn.microsoft.com/en-us/sql/t-sql/functions/rank-transact-sql?view=sql-server-ver16

It seems these other ones would have ties take the lower values, whereas for arrow-rs the existing rank function (and subsequently what is introduced here) takes the higher value; I guess this might be worth a broader discussion, as this PR itself currently follows the existing behaviour

:thinking:

_Originally posted by @Jefffrey in https://github.com/apache/arrow-rs/pull/6912#pullrequestreview-2561770071_

Start by reviewing the existing rank behavior in arrow-rs and the referenced discussion in pull request #6912. Before changing anything, establish the intended tie-ranking semantics against PostgreSQL, DuckDB, and SQL Server, then add or update tests that demonstrate the agreed behavior.

Written by the indexing model from the issue text.