Simplify package installation for developing and running expectations

[mateocolina]

Is your feature request related to a problem? Please describe.
When developing expectations, one can use multiple ways to do so. Whether you would like to use JupyterLab is up to you. For running expectations in GX, a lot of packages are not required and it blows the environment (local, Docker, K8s etc.) unnecessarily (last I checked after installing GX with acryl-datahub, the Docker image had a size of 1.1 GB). We distinguish between developing expectations and running them in our daily pipeline, thus the setup is also different.

Describe the solution you'd like
If running pip install great-expectations it should install all packages related to GX. But pip install "great-expectations[ops]" should only install packages to successfully run expectations in the environment.

Describe alternatives you've considered
Uninstalling packages after the Docker image has been built, but with each release the packages are installed again.

Additional context
Add any other context, screenshots, or code samples about the feature request here.