fix: TopkDropoutStrategy honor overridden get_risk_degree() for buy sizing

[zhaow-de]

Description

TopkDropoutStrategy.generate_trade_decision sized new buys with the raw attribute self.risk_degree instead of the getter self.get_risk_degree(trade_step) (qlib/contrib/strategy/signal_strategy.py:266):

# before
value = cash * self.risk_degree / len(buy) if len(buy) > 0 else 0
# after
value = cash * self.get_risk_degree(trade_step) / len(buy) if len(buy) > 0 else 0

This was the only use of risk_degree in TopkDropoutStrategy, and it bypassed the getter. The change makes Topk consult get_risk_degree() for buy sizing, matching WeightStrategyBase (line 365) and its subclass EnhancedIndexingStrategy (line 482), which already call self.get_risk_degree(...). It is behavior-preserving for the default case, since BaseSignalStrategy.get_risk_degree() returns self.risk_degree by default.

Motivation and Context

Related issue: closes #2276.

get_risk_degree() is documented as the dynamic market-timing hook ("Dynamically risk_degree will result in Market timing.", signal_strategy.py:69). Because Topk read the raw attribute instead of the getter, a subclass overriding get_risk_degree() to return a time-varying multiplier (the documented way to implement market timing) had zero effect on a TopkDropout book — the override was never consulted for position sizing. Two strategies overriding the hook would therefore produce identical trades/returns. This fix makes the documented override mechanism actually work for TopkDropoutStrategy.

How Has This Been Tested?

• Pass the test by running: pytest qlib/tests/test_all_pipeline.py under upper directory of qlib.
• If you are adding a new feature, test on your own test scripts.

Added a regression test at tests/backtest/test_topk_risk_degree.py:

• test_gate_to_cash_via_get_risk_degree — a TopkDropoutStrategy subclass that overrides get_risk_degree() to return 0.0 (gate to cash) now builds no stock position. Fails before this change (it bought stock using the raw 0.95 default), passes after.
• test_base_strategy_does_buy — sanity check that the default strategy still buys.

Screenshots of Test Results (if appropriate):

1. Pipeline test (tests/test_all_pipeline.py, default-risk-degree TopkDropoutStrategy backtest — confirms the default path is unchanged):

test_all_pipeline.py::TestAllFlow::test_0_train PASSED                   [ 33%]
test_all_pipeline.py::TestAllFlow::test_1_backtest PASSED                [ 66%]
test_all_pipeline.py::TestAllFlow::test_2_expmanager PASSED              [100%]
================== 3 passed, 3 warnings in 247.39s (0:04:07) ===================

2. Your own tests (tests/backtest/test_topk_risk_degree.py):

backtest/test_topk_risk_degree.py::TopkRiskDegreeTest::test_base_strategy_does_buy PASSED [ 50%]
backtest/test_topk_risk_degree.py::TopkRiskDegreeTest::test_gate_to_cash_via_get_risk_degree PASSED [100%]
======================= 2 passed, 21 warnings in 27.09s ========================

Types of changes

• Fix bugs