Make bi-objective NDS optimization opt-in and move benchmark to tests #754

- Add native_biobj_sorting parameter to fast_non_dominated_sort (default False)
- Move test_comparison.py from pymoo/ to tests/benchmark/run_native_biobj.py
- Simplify benchmark to use actual pymoo implementation

Benchmark results show minimal performance improvement:

  Size       Default (ms)    Native BiObj (ms)  Speedup    Correct
  50         0.107           0.036              2.98x      Yes
  100        0.103           0.100              1.03x      Yes
  500        1.038           1.065              0.98x      Yes
  1000       3.205           3.094              1.04x      Yes
  2000       8.615           8.642              1.00x      Yes

Not much improvement compared to moocore, thus not enabled by default.

Addresses review comments from @MLopez-Ibanez

Author: blankjul

pymoo/functions/standard/non_dominated_sorting.py

@@ -10,19 +10,30 @@
 from pymoo.util.dominator import Dominator
 
 
-def fast_non_dominated_sort(F, dominator=Dominator(), **kwargs):
-    """Fast non-dominated sorting algorithm."""
+def fast_non_dominated_sort(F, dominator=Dominator(), native_biobj_sorting=False, **kwargs):
+    """Fast non-dominated sorting algorithm.
+
+    Parameters
+    ----------
+    F : np.ndarray
+        Objective values for each individual.
+    dominator : Dominator
+        Dominator object for custom dominance relations.
+    native_biobj_sorting : bool
+        If True, use specialized O(N log N) algorithm for bi-objective problems.
+        Default is False.
+    """
     if F.size == 0:
         return []
-    
+
     n_points, n_objectives = F.shape
-    
+
     # For single objective or single point, return immediately
     if n_points <= 1:
         return [list(range(n_points))] if n_points == 1 else []
-    
-    # For bi-objective problems, use specialized O(N log N) algorithm
-    if n_objectives == 2:
+
+    # For bi-objective problems, optionally use specialized O(N log N) algorithm
+    if native_biobj_sorting and n_objectives == 2:
         return _fast_biobjective_nondominated_sort(F)
 
     if "dominator" in kwargs: