All tools should make use of public runtime container image to manage dependencies

We've finally been able to host a public Docker image for weather-tools. The first version can be pulled with this command:

docker pull gcr.io/weather-tools-prod/weather-tools:0.0.0

This unblocks our ongoing efforts to make weather-tools easy to install!

In brief, here is what's needed to make installation better:

For each tool's associated script, let's add an argument that points to our runtime container. Right now, we specify extra_packages. I believe this can go away. See https://github.com/google/weather-tools/blob/edd742d591cb428aa41f6c4d713c969e39f8dbbe/weather_dl/weather-dl#L76. I believe we can remove a lot of the packaging and installing logic in these scripts.
We will most likely need to create a new runtime container with an incremented version to get the latest dependencies.

The next steps to fully land a clean install experience will involve:

Landing the conda install google-weather-tools experience via this staged recipe: https://github.com/conda-forge/staged-recipes/pull/20892
Following our release process and tagging a new public release.
Setting up CI to keep our conda environment and docker image versions consistent via Google Cloud Build and GCR.

CC: @blackvvine @DarshanSP19

google / weather-tools