Embedder performance testing

Write load test that sends a short sentence in the following payload:

sentences = ["The sky is blue."]

data = {"data": {"ndarray": sentences}}

To the following 2 endpoints:

https://ml-staging.delvin.to/api/v1.0/predictions
https://ml2-staging.delvin.to/api/v1.0/predictions

ml2-staging is a new embedder we're trying out, and trying to understand if it's significantly slower than the smaller model.