Accelerate data retrieval pathways in the Cursor class by optimizing row
iteration and reducing interpreter overhead on high-frequency method loops.

Detailed changes:
- fetchone: Inline the `_check_closed()` state verification directly into the
  method body. This removes function-call stack overhead for every individual
  row retrieved.
- fetchall: Refactor the loop to slice and drain the current in-memory result
  batch in bulk using `list.extend()` instead of sequentially invoking
  `fetchone()` row by row.
- fetchall: Directly request the next network payload batch via
  `fetch_result_set_next` once the in-memory collection is exhausted, updating
  total `rownumber` tracking as a single block operation.
- fetchall: Add a defensive guard clause against unallocated result sets to
  ensure strict DB-API compatibility.