PR for Issue #9740

This PR performs a broad cleanup of the statsmodels codebase to improve maintainability and addresses a specific corner case in Kalman filter predictions. Key changes include removing dead code, implementing missing convergence warnings, and refining diffuse initialization logic.

Detailed Changes

1. Bug Fixes & Improvements

• Kalman Filter (kalman_filter.py): Fixed a corner case where dynamic/out-of-sample predictions were incorrectly initialized if the filter hadn't exited the diffuse phase. Now uses Initialization.from_components with the diffuse covariance when predicted_diffuse_state_cov is non-zero.
• SARIMA Warnings (innovations.py): Added ConvergenceWarning to innovations_mle when the nonlinear optimizer fails (success=False).

2. Dead Code Removal (Library)

• regressionplots.py: Removed internal helper _partial_regression which was unused and redundant.

3. Test Suite Cleanup

• test_glm.py: Removed extensive commented-out test blocks relying on obsolete rpy2 infrastructure.
• test_regression.py: Removed multiple unused check_confidenceintervals methods and associated FIXME comments.
• test_var.py: Removed the unused CheckVAR class and broken generate_var functions.
• test_data.py: Removed the commented-out TestDates class.

4. Documentation & Annotation

• kernels.py: Removed a stale FIXME in aitchison_aitken_cdf. The logic was verified to be correct for joint CDF definitions of encoded discrete variables.

Verification Plan

• Manual Inspection: Verified that all removed code has no callers in the current repository.
• State Space Logic: Verified that predict now correctly passes Pinf when initializing the forecast model during the diffuse phase.
• Local Testing: Individual syntax and logic checks were performed; full test suite execution was limited due to environment/versioning constraints with pytest.