EPIC: use cp_solver framework to develop a more sophisticated predicate simplification

### Is your feature request related to a problem or challenge?

The predicates in filter can be variable, it's possible to use the cp_solver to simplify the predicates in filter, then reduce the computing cost for filter.

### Describe the solution you'd like

We've done some basic simplification in https://github.com/apache/datafusion/pull/16362, we can make it a start to develop a more sophisticated rule to cover more cases

### Describe alternatives you've considered

_No response_

### Additional context

What I want to discuss:
1. Make it a new physical optimizer rule? Or add them to the current physical filter push down
2. Some filters, such as partition columns related predicates, will be fully pushed down scan, that is, we don't have such predicates in some FilterExec. How to capture this?
3. TBD (You may have other thoughts!)

Start by reviewing the basic predicate simplification in pull request 16362 and the physical filter pushdown path, including FilterExec. The work first requires deciding whether this belongs in a new physical optimizer rule or existing filter pushdown, and how to handle predicates fully pushed into scans; done means an agreed approach that covers more simplification cases.

Written by the indexing model from the issue text.