Comment out QSVC/VQC and make retrain optional in quantum classification example

[Copilot]

The quantum classification example failed at runtime, and manually commenting out the retrain(...) call left epochs / error_rate undefined further down. QSVC and VQC baselines also need to be disabled for now.

Changes to examples/quantum/classification.py

• Retrain toggle: introduced RETRAIN, RETRAIN_EPOCHS, RETRAIN_LR module constants. The retrain call is wrapped in if RETRAIN: ... else: ...; the else branch assigns safe defaults (error_rate=nan, epochs=0, fold_epoch_history=[]) so the checkpoint payload and downstream int(epochs) / float(error_rate) remain valid whether or not retrain runs. lr is now passed explicitly instead of relying on the model default.
• QSVC / VQC disabled: commented out qiskit imports, per-fold QSVC and VQC training/eval blocks, their keys in the fold_data dict, the __main__ aggregation lists, print_cv_summary calls, ROC data prints, and roc_curve plot-point sections. Inline notes indicate exactly what to uncomment to re-enable.

RETRAIN = True
RETRAIN_EPOCHS = 10
RETRAIN_LR = 1.0

if RETRAIN:
    error_rate, epochs = model_q.retrain(
        X_train_fold, y_train_fold,
        epochs=RETRAIN_EPOCHS, lr=RETRAIN_LR, verbose=False,
    )
    fold_epoch_history = list(getattr(model_q, "retrain_history_", []))
else:
    error_rate = float("nan")
    epochs = 0
    fold_epoch_history = []

Flip RETRAIN = False at the top of the file to run the example without the retrain step; no other edits required.