feat(query): Introduced fold_for_prune to optimize index prune

[forsaken628]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

For any injective function y = f(x), the eq(column, inverse_f(const)) is a necessary condition for eq(f(column), const).
For example, the result of col = '+1'::int contains col::string = '+1'.

part of #13408

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test - Explain why

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

---

This change is 

[sundy-li]

I did not see any codes to match the comment.

[forsaken628]

This comment means that Ignore FunctionDomain::MayThrow is a precondition for is_injective_cast. It will be refined later.

[sundy-li]

need to check that string literal matches [1-9][0-9]*

[sundy-li]

should be false.

[forsaken628]

fold_for_prune is similar to bloom-filter in that the expressions generated are all necessary but not sufficient conditions.

In fact most of the above tests are necessary but not sufficient, e.g. the sufficient condition for to_int8(bool) = 2 is false because the result of the original expression is the empty set, and only the empty set is a subset of the empty set.

[b41sh]

It looks like fold_for_prune is meant to rewrite the equal expression with a cast column function, like col::dest_type = val as col = val::src_type, which we can rewrite directly inside the visit_expr_column_eq_constant function and don't need redefine fold_for_prune function. If val::src_type fails, the result must be false.