A2Q+: Improving Accumulator-Aware Weight Quantization