Refactor `SortExec`'s buffered batches for better code readability

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

Reference https://github.com/apache/datafusion/pull/15355#discussion_r2009077643 https://github.com/apache/datafusion/pull/15355#discussion_r2009002846

The high-level execution logic of `SortExec` is described in https://github.com/apache/datafusion/blob/main/datafusion/physical-plan/src/sorts/sort.rs, now a single field `in_mem_batches` is used to represent buffered data in different stage: during different time, it can be interpreted as either unordered input batch, or globally sorted batch:
https://github.com/apache/datafusion/blob/5210a2bac32e43dc7bf6e7e6000cdeaf2833c06e/datafusion/physical-plan/src/sorts/sort.rs#L229
This approach has poor understandability and is also error-prone, see the idea in reference discussion for improvements.

### Describe the solution you'd like

_No response_

### Describe alternatives you've considered

_No response_

### Additional context

_No response_