Add plan_any/plan_all to ExprPlanner to decouple functions from sql crate

[Jefffrey]

Which issue does this PR close?

• N/A

Rationale for this change

sql crate toml states that it should not depend on the functions crates:

datafusion/datafusion/sql/Cargo.toml

Lines 49 to 59 in 127731b

	# Note the sql planner should not depend directly on the datafusion-function packages
	# so that it can be used in a standalone manner with other function implementations.
	#
	# They are used for testing purposes only, so they are in the dev-dependencies section.
	[dependencies]
	arrow = { workspace = true }
	bigdecimal = { workspace = true }
	chrono = { workspace = true }
	datafusion-common = { workspace = true, features = ["sql"] }
	datafusion-expr = { workspace = true, features = ["sql"] }
	datafusion-functions-nested = { workspace = true, features = ["sql"] }

However on line 59 it is still depending on the nested functions crate, due to any/all planning support:

datafusion/datafusion/sql/src/expr/mod.rs

Lines 1263 to 1272 in 127731b

	/// Plans a `<left> <op> ANY(<right>)` expression for non-subquery operands.
	fn plan_any_op(
	left_expr: Expr,
	right_expr: Expr,
	compare_op: &BinaryOperator,
	) -> Result<Expr> {
	match compare_op {
	BinaryOperator::Eq => Ok(array_has(right_expr, left_expr)),
	BinaryOperator::NotEq => {
	let min = array_min(right_expr.clone());

• Since needs access to array_has function. Aim to decouple these crates via using ExprPlanner trait.

What changes are included in this PR?

Introduce plan_all and plan_any to ExprPlanner, and move the implementation code into datafusion-functions-nested, removing dependency of datafusion-functions-nested from datafusion-sql

Are these changes tested?

Existing tests.

Are there any user-facing changes?

Hopefully not? Theres new methods on ExprPlanner but they have default implementations, and if using NestedFunctionPlanner (which should be there by default I believe) it'll still plan any/all correctly.

[Jefffrey]

this implementation code is copied essentially verbatim from sql/src/expr/mod.rs; the only modifications being some renaming of variables (for any it called them left/right which are now aligned to needle/haystack) and some plumbing to make it work with the signatures of plan_any/plan_all

[Jefffrey]

main benefit here

[Jefffrey]

also moving this planning code into datafusion-functions-nested should allow us to more easily create dedicated UDFs for any/all without needing to expose these UDFs beyond the crate

• e.g. instead of needing to create new UDFs array_has_all_eq/array_has_all_gt that must be public for the planning code to reach it (from sql/src/expr/mod.rs) we can keep it private within this crate

related:

• Replace ANY/ALL CASE planning with array_has/min/max desugaring #22102

[Jefffrey]

one part i'm not sure about was whether to include this subquery case as part of this plan_any/plan_all call; maybe its better to just name these as plan_any_nested/plan_all_nested to differentiate it from the subquery planning, which we'll keep in datafusion-sql?