[BUG] Fix DummyProbaRegressor._predict_var returning std dev instead of variance

[maniktyagi04]

Fixes #975

Summary

DummyProbaRegressor._predict_var was incorrectly returning the standard deviation (σ) of training labels instead of the variance (σ²). This PR fixes the bug and adds a dedicated test file.

---

Changes

skpro/regression/dummy.py

In _fit: Added storage of self._var = np.var(y.values) alongside the existing self._sigma.

# Before (buggy)
self._sigma = np.std(y.values)

# After (fixed)
self._sigma = np.std(y.values)
self._var = np.var(y.values)   # ← NEW: store variance explicitly

In _predict_var: Changed to use self._var instead of self._sigma.

# Before (buggy)
np.ones(X_n_rows) * self._sigma   # returned std dev σ

# After (fixed)
np.ones(X_n_rows) * self._var     # returns variance σ²

---

skpro/regression/tests/test_dummy.py (NEW)

Added 5 targeted tests for DummyProbaRegressor.predict_var:

Test	What it checks
test_predict_var_equals_training_variance_empirical	Returns np.var(y_train) for 'empirical' strategy
test_predict_var_equals_training_variance_normal	Returns np.var(y_train) for 'normal' strategy
test_predict_var_not_equal_std	Regression: does NOT return std dev
test_predict_var_is_nonnegative	Variance is always ≥ 0
test_predict_var_index_and_columns	Output index/columns match X_test/y_train

All 5 tests pass locally.

---

How to verify

python -m pytest skpro/regression/tests/test_dummy.py -v

Expected: 5 passed

---

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New tests