normen662

In the Cascades planner we try to reuse references if possible. We should not ever reuse references that can be pruned later on.

change modeling of `OrderingPart` to use a sort order instead of a boolean for reversedness

1

This is related to #1318. It would be useful to state that certain ordering parts are fixed, i.e. bound to one value vie an e.g. equality predicate. In those cases...

### Example 1 We have an index on `15 - y`, it can be used for a query predicate `y < 10`! Something should turn `minus(15, y)` into `f(y)` and...

This issue is to exclusively use `IndexKeyValueToPartialRecord` in the heuristic planner and use a `RecordConstructorValue` or similar to compute the partial record that the `RecordQueryCoveringIndexScanPlan` can then return.

In #2858 we kept some code in `ScanWithMatchCandidate` that allowed us to create compatible plans when not planning mixed order ordering requirements. This should be removed once downstream is adapted.

This logic should be refactored in a way that it calls the implies-logic in PredicateWithValuesAndRanges instead of replicating that logic here. The only reason why that is not done right...