You've got in fact three questions here. If they are all related, then you are probably interested in getting datasets of thermal images of mango plantations, e.g. from a drone or airplane. I doubt that there are datasets for that.
If you are interested in mango fruits, like on a dish, then you can write a web-crawler to scrap Google Images or other websites for keyword "mango". Such crawlers are usually created using Scrapy library in Python language. There are also many available code for scraping Google images, you can easily find them on the web. This will provide you with mango fruit images, but most likely not in the IR spectrum (if that's what you need at all). If you need images in a specific spectrum (e.g., thermal) then the most appropriate solution would be to create this dataset by taking pictures yourself...
Regarding how thermal image processing works in agriculture - I'm not an expert in agriculture but I know of some applications like taking pictures of vineyards from a drone and then find out the regions that are hotter than others (it's usually very visible in the thermal images) and direct water to only those regions. This saves water resources.