Run the WASM ML inference model in k3s

[jiayihu]

Epic #51

• [x] Allow to reach the WASM Pod from the cluster #58
• [x] Run river flow ML model in krustlet #59

[jiayihu]

The idea of the application architecture is the following:

• The Flood application receives rain measurements from the rainfall sensor exposed as service and it combines it with the measurement from the day before and the average flow of the day before
• A service allows to run the inference on a krustlet running the ONNX inference model exposes as HTTP service
• At the same time, the data is aggregated and sent to a GCE cloud database at the end of the day, so that training can be run

The first benchmark will test the response time of the inference from the moment the data is received from the sensor and the inference is requested. Later, another benchmark could show how quickly the cluster can spawn a service in case of interruption.