Refactor image conversion (TIFF->PNG) and thumbnail code

[jreyesr]

The code that handles the conversion of orthomosaics and DEMs from TIFF to PNG and the PNG thumbnail creation is currently (relatively) brittle, it uses different code paths for RGB and Micasense images, and it breaks on really big GeoTIFFS.

• Image thumbnail creation could use the code at https://stackoverflow.com/q/50465366 (gdal_translate -of PNG -outsize 512 512 odm_orthophoto.tif uuid_thumbnail.png, place the thumbnail on the tmp folder)
• TIFF to PNG conversion could use something like https://stackoverflow.com/q/61764260
• Remember to preserve the following lines (though only on a multispectral flight!) https://github.com/IS-AgroSmart/AgroSmart-Web/blob/e5928c976127e68bfcbc284ed81d5867331933d9/core/models.py#L225-L226

Any changes must be tested on:

• An orthomosaic from a normal, sane RGB flight
• An orthomosaic from a multispectral camera
• An orthomosaic from a really big RGB flight (current code fails on a 1.6GB mosaic)

The results to test are:

• PNG thumbnail must be created (512x512 or something similar) and must be saved on the tmp folder where the frontend code can reach it.
• PNG orthomosaic must be created and readable by an image viewer
• If Micasense camera was used, a file called rgb.tif must have been created and must load correctly into Geoserver (check the bands and transparency)