Add more examples using different modalities for the image classification task

[xiong-zhitong]

Add fine-tuning examples on the Sentinel-1, Sentinel-2, and hyperspectral datasets; Add the usages for online demo based on the Gradio.