[WIP] core: group reduction using union find

[yliang412]

The PR uses union-find to maintain group equivalence. This is done by maintaining

• a DisjointSet of group ids
• representative group id to Group mapping.

The "path compression" done during union operation speeds up tpch sqlplannertest run from ~89s to ~9s on local machine.

TODO

• Better to have a Map<GroupId, Group> interface without thinking about reduced group. We can use DisjointGroup to replace get_reduced_group_id, Memo::groups::..., and Memo::merge_group functionalities.