[CHORE] Use treenode for tree traversal in logical optimizer rules

[kevinzwang]

Follow-up from #2791 after offline conversation.

This PR still uses the existing logical plan optimizer, but now uses common_treenode::Transformed instead of a custom implementation. In addition, it removes the apply order logic from the optimizer and replaces it with common_treenode::TreeNode transforms done in the optimizer rules themselves.

This PR should not cause any functional changes to any of the optimizer rules or the way they are applied, except for the fact that rules in a batch are now each applied to the whole tree before the next rule, instead of each being applied to a single plan node if they have the same apply order. Future PRs will separate the rules out and make better use of treenode.

[codspeed-hq]

CodSpeed Performance Report

Merging #2797 will degrade performances by 12.25%

_{Comparing kevin/logical-optimizer-treenode (c9ea3bb) with main (6fe408c)}

Summary

⚡ 1 improvements
❌ 1 regressions
✅ 14 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	kevin/logical-optimizer-treenode	Change
❌	test_count[1 Small File]	20.9 ms	23.8 ms	-12.25%
⚡	test_show[100 Small Files]	59.5 ms	51.9 ms	+14.56%

[samster25]

why do we need this clippy allow? This should be used right?

[kevinzwang]

&self is not actually used other than recursive invocations. It's weird because I don't think Clippy complains if you don't use &self at all, only if there's a recursive call

[kevinzwang]

This is just like a quick and dirty fix to make Clippy happy and ensure that things stay the same before we start actually ripping into the rules themselves.