fix(ops): guard Rsquare expanding (N==0) path against degenerate windows

[DogInfantry]

Fixes #2297

Problem

Rsquare masks windows whose std is ≈0 to NaN so a degenerate 0/0 regression
(near-constant window) does not leak inf/garbage R² out of the Cython kernel.
That guard was only on the rolling (N != 0) branch; the expanding (N == 0)
branch was unguarded, so Rsquare($feature, 0) returned inf/spurious finite
values on near-constant windows. qlib/data/ops.py also sets
np.seterr(invalid="ignore"), so the bad values propagate silently into
features (Alpha158/Alpha360) and models.

Fix

Mirror the existing guard on the expanding branch, using expanding std:

if self.N == 0:
    series = pd.Series(expanding_rsquare(_series.values), index=_series.index)
    series.loc[np.isclose(_series.expanding(min_periods=1).std(), 0, atol=2e-05)] = np.nan
else:
    ...

Slope/Resi are unaffected — they divide by the x-variance (index 1..N),
which is always well-conditioned; only Rsquare divides by the y-variance.

Test

Adds tests/ops/test_rsquare.py exercising Rsquare._load_internal directly
(no market data needed). It fails before this change and passes after:

# before fix, expanding path:
[nan, nan, nan, inf, 0.01717987, inf]   -> AssertionError: expanding Rsquare leaked inf
# after fix:
3 passed

[DogInfantry]

@microsoft-github-policy-service agree