feat: optimise outer joins

[systay]

Description

This PR introduces enhancements to the query planner, enabling the push down of LIMIT clauses to the right-hand side (RHS) of joins, including outer joins, when the expressions permit. Additionally, logic has been added to push down LIMIT under routes when applicable.

These changes aim to optimize query efficiency and resource utilization, ensuring faster query processing and lower overhead in the presence of outer joins.

For inner joins, we can't push down LIMIT on the left-hand side (LHS) of our nested loop joins (ApplyJoin). For a join to produce a row, we need a match on both sides, and we don't know how many rows we need from the LHS for the join to produce sufficient rows to satisfy the user-set LIMIT. The best we can do is push a LIMIT on the RHS and then also LIMIT on top of the join.

On the other hand, for left outer joins, we know that every row from the LHS will be returned from the join. This means we can push down the LIMIT to the LHS of the join. This is really good and important - we want to run the RHS of the join as few times as possible.

Example

SELECT * 
FROM user 
  JOIN user_extra ON user.col = user_extra.col
LIMIT 10

The first, unoptimized version of the query with the LIMIT looks like this:

Limit
└── ApplyJoin
    ├── Route
    │   └── Table (user.user)
    └── Route
        └── Filter (joinCondition)
            └── Table (user.user_extra)

Now we rewrite by doing iterative tree rewrites until we reach a fixed point. It will look something like this:

Limit (1)
└── ApplyJoin
    ├── Limit (2)
    │   └── Route
    │       └── Limit (3)
    │           └── Table (user.user)
    └── Limit (4)
        └── Route
            └── Limit (5)
                └── Filter (joinCondition)
                    └── Table (user.user_extra)

Here are the reasons for the different LIMIT operations:

1. This is the top LIMIT. It ensures that the final results do not contain more than 10 rows.
2. We want to run as few rows through our ApplyJoin as possible, so if the Route under is not a single-sharded Route, we add a LIMIT on top of the LHS of the JOIN.
3. We know we don't need more than 10 rows from each shard, so we can push down the LIMIT under the route. 🎉
4. Similar to (2) - if the route under is touching multiple shards, we keep a LIMIT on top of the Route to minimize the number of rows processed by the join.
5. Finally, we know we don't need more than 10 rows from any shard on the RHS of the join.

Related Issue(s)

Fixes #15828

Checklist

• "Backport to:" labels have been added if this change should be back-ported to release branches
• If this change is to be back-ported to previous releases, a justification is included in the PR description
• Tests were added or are not required
• Did the new or modified tests pass consistently locally and on CI?
• Documentation was added or is not required**

[vitess-bot]

Review Checklist

Hello reviewers! 👋 Please follow this checklist when reviewing this Pull Request.

General

• Ensure that the Pull Request has a descriptive title.
• Ensure there is a link to an issue (except for internal cleanup and flaky test fixes), new features should have an RFC that documents use cases and test cases.

Tests

• Bug fixes should have at least one unit or end-to-end test, enhancement and new features should have a sufficient number of tests.

Documentation

• Apply the release notes (needs details) label if users need to know about this change.
• New features should be documented.
• There should be some code comments as to why things are implemented the way they are.
• There should be a comment at the top of each new or modified test to explain what the test does.

New flags

• Is this flag really necessary?
• Flag names must be clear and intuitive, use dashes (-), and have a clear help text.

If a workflow is added or modified:

• Each item in Jobs should be named in order to mark it as required.
• If the workflow needs to be marked as required, the maintainer team must be notified.

Backward compatibility

• Protobuf changes should be wire-compatible.
• Changes to _vt tables and RPCs need to be backward compatible.
• RPC changes should be compatible with vitess-operator
• If a flag is removed, then it should also be removed from vitess-operator and arewefastyet, if used there.
• vtctl command output order should be stable and awk-able.