Remove integer indexing from game collections in pygambit

[d-kad]

Description

Game object collections in pygambit are now indexed by string label only.

• Integer indexing is removed from Game.players, Game.outcomes, Game.strategies, Game.infosets,
  Game.actions, Player.strategies, Player.infosets, Player.actions, Infoset.actions, Infoset.members,
  and Node.children. Indexing by an integer raises TypeError.
  Positional access remains available by iterating a collection (e.g. pl1, pl2 = game.players).
  Indexing a Game by a contingency (e.g. game[0, 1]) is unchanged.
• Label lookup is by exact match.
  A label matching no object uniformly raises KeyError (Node.children previously raised ValueError).
• Lookup is centralised in a new _resolve_by_label helper; all the collection __getitem__ methods use it.
  The label parameter is type-hinted str, so Cython rejects non-string indices at the call boundary.
• The test suite (including the tutorial notebooks) is fully migrated to label-based access (~460 affected tests),
  with new tests added pinning the integer-rejection across all collections
  (COLLECTION_GETTERS in tests/test_game.py).

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[d-kad]

Design question — should label lookup strip whitespace, or match exactly?

Before this PR, the collection __getitem__s stripped the lookup string before comparing:

matches = [x for x in self if x.label == index.strip()]

This appears at 10 sites (game.pxi ×5, player.pxi ×3, infoset.pxi ×2),
so e.g. g.players[" Alice "] would find the player labelled "Alice".
The most recent new Node.children semantics in #588, matches labels verbatim.

In this PR I matched labels exactly. ~~ (superseded — see update)

UPD: @rahulsavani and I agreed label lookup should keep stripping surrounding
whitespace from the query (stored labels are matched as-is).

[tturocy]

A few places where we should be using Cython type hints.