If there are duplicate members when joining with the users
table, it results in a lot of unnecessary memory usage.

The members table can contain the same user_id multiple times
if the user_id is part of multiple groups/projects. Currently, the
SQL query generated by Group#billed_project_members joins the
members and users tables. The duplicate user_ids in the members
table can result in unnecessary memory usage.
This commit modifies Group#billed_project_members to filter out
duplicate members rows first in a subquery, and the subquery is then
used as an IN condition to query the users table.

It also renames Group#billed_project_members to
Group#billed_project_users.