[Quantization] Support NVFP4 for inline-swiglu fused MoE experts (MiniMax-M3 / MiniMaxM3VLExperts)

[Quantization] Support NVFP4 for inline-swiglu fused MoE experts (MiniMax-M3 / MiniMaxM3VLExperts) #5292