This is a proof-of-concept for implementing the proxy based download statistics

Basically, we can change a model weight download url like this:

https://bioimage.netlify.app/.netlify/functions/download/10.5281/zenodo.5764892/files/weights.onnx?debug=1

(with ?debug=1, it will return the the generated url but not sending request, for actual use we can remove it)

This will return something like:

{"actualFileUrl":"https://zenodo.org/api/records/5764892/files/weights.onnx/content","matomoReportUrl":"https://bioimage.matomo.cloud/matomo.php?download=https://doi.org/10.5281/zenodo.5764892&idsite=1&rec=1&r=646242&h=13&m=35&s=20&url=http://bioimage.io/#/?id=10.5281/zenodo.5764892&uadata=%7B%22brands%22%3A%5B%7B%22brand%22%3A%22unknown%22%2C%22version%22%3A%22unknown%22%7D%5D%7D"}

The actualFileUrl will be send for download, meaning the user will be redirected to the actual URL, and matomoReportUrl will be used for reporting a model download for a model. If you actually visit the matamoReportUrl, you will see the report here in realtime: https://bioimage.matomo.cloud/

We can also get the model statistics via API, @FynnBe has done some work here: bioimage-io/spec-bioimage-io#504

cc @FynnBe @ctr26