Is your feature request related to a problem? Please describe.
Our team (Kaggle) supports a Public Datasets BigQuery integration in Kernels (Jupyter Notebooks) using a proxy where we inject our own credentials. We currently expose a PublicBigQueryClient that handles this (https://github.com/Kaggle/docker-python/blob/master/patches/kaggle_gcp.py#L70) but we're unable to support Magics via this mechanism since we need to be able to pass in a Connection object to the client.
Describe the solution you'd like
Being able to set either the client object or the Connection object of the Magic (like Context.set_credentials) would solve this.
Describe alternatives you've considered
We've considered monkeypatching the magics client but that may cause more problems.