Introduce new ways to split an automaton

The explicit way of splitting suffers if there are
too many input APs, two new ways of splitting
are introduced as well as a heuristic to chose
between them.

* NEWS: update
* spot/twaalgos/synthesis.cc,
spot/twaalgos/synthesis.hh: New fonctions
* bin/ltlsynt.cc: Add corresponding option
* tests/core/gamehoa.test,
tests/core/ltlsynt.test,
tests/python/_partitioned_relabel.ipynb,
tests/python/_synthesis.ipynb,
tests/python/game.py,
tests/python/split.py,
tests/python/synthesis.py: Adjusting and adding test

tests/python/synthesis.py

@@ -29,8 +29,7 @@ for i in range(0, 2):
     tc.assertFalse(spot.solve_game(game))
 
 # A game can have only inputs
-game = spot.ltl_to_game("GFa", [])
-tc.assertEqual(game.to_str(), """HOA: v1
+game_ref = spot.automaton("""HOA: v1
 States: 3
 Start: 0
 AP: 1 "a"
@@ -49,3 +48,10 @@ State: 1
 State: 2
 [t] 0 {0}
 --END--""")
+
+gi = spot.synthesis_info()
+gi.dict = game_ref.get_dict()
+
+game = spot.ltl_to_game("GFa", [], gi)
+
+tc.assertTrue(spot.are_equivalent(game, game_ref))