docs: Document that adding new optimizer rules are expensive

[alamb]

Which issue does this PR close?

• Similarly to Update comments on OptimizerRule about function name matching #20346

Rationale for this change

As part of PR reviews, it seems like it is not obvious to some contributors that there is a non trivial cost to adding new optimizer rules. Let's add that knowledge into the codebase as comments, so it may be less of a surprise

What changes are included in this PR?

Add comments

Are these changes tested?

N/A

Are there any user-facing changes?

No this is entirely internal comments oly

[adriangb]

Left a small suggestion

[adriangb]

Suggested change

	// - Adding a new rule here is expensive as it will be applied to all
	// queries, and will likely increase the optimization time. Please extend
	// existing rules when possible, rather than adding a new rule.
	// - Adding a new rule here is expensive as it will be applied to all
	// queries, and will likely increase the optimization time. Please extend
	// existing rules when possible, rather than adding a new rule.
	// If you do add a new rule considering having aggressive no-op paths
	// (e.g. if the plan doesn't contain any of the nodes you are looking for
	// return `Transformed::no`; only works if you control the traversal).