[EPIC] Support nondetermenistic expressions

Propose adding native implementations for Spark SQL's nondeterministic functions:
```
monotonically_increasing_id()
rand()
randn()
spark_partition_id()
```

The first step is to refactor `from_proto.rs` into `planner.rs` as `PhysicalPlanner`, enabling it to support the passing of `SessionContext` and `TaskContext` during the conversion of Spark query plans to DataFusion query plans.


## Sub Tasks

- [x] #1871
- [x] #1888
- [x] #1889
- [ ] #1890
- [x] #1891