Fix analysis engine receiving bare 'go' instead of 'go infinite'

[Copilot]

Review mode analysis engine sits idle because it receives the UCI command go (no parameters) instead of go infinite. The board never updates and the GUI becomes unresponsive.

Root cause

python-chess's engine.analysis(board, limit) decides whether to append infinite via if limit:. A Limit() dataclass with all-None fields is truthy, so passing Limit(depth=None) produces bare go instead of go infinite:

# python-chess internals (engine.py)
if limit:
    self.engine._go(limit, root_moves=root_moves)          # sends "go"
else:
    self.engine._go(Limit(), root_moves=root_moves, infinite=True)  # sends "go infinite"

Changes

• Pass limit=None for infinite analysis — When tc_type='infinite' with no depth constraint, pass None instead of Limit(depth=None) so python-chess takes the else branch and sends go infinite. Custom depth limits still use Limit(depth=X) → go depth X.

# Before: Limit(depth=None) is truthy → bare "go"
limit = chess.engine.Limit(depth=self.max_depth if self.max_depth != MAX_DEPTH else None)

# After: None is falsy → "go infinite"
limit = (chess.engine.Limit(depth=self.max_depth)
         if self.max_depth != MAX_DEPTH else None)

• Guard engine.play() fallback against None limit — The bestmove fallback path now also checks limit is not None to avoid passing None to engine.play().

• Skip engine.play() fallback when killed (prior commit) — After _kill.set(), RunEngine.run() was falling through to engine.play(board, Limit(depth=None)) which blocked the thread indefinitely. Now skipped when the kill signal is set.

• Non-blocking review navigation with debounce (prior commit) — Navigation signals the analysis thread to stop without blocking on join(). The actual restart is deferred via a 300ms debounce in the timeout handler, keeping the GUI responsive during rapid navigation.