Limitations of using shared mcr.microsoft.com/mlops/python:latest container

[algattik]

• The steps for adapting the pipeline to use other ML modules than scikit-learn or additional Python modules besides pandas (provided in the sample Dockerfile and requirements.txt) are not obvious.
• While iterating on such ML modules or on their package versions, there is interference between branches when using latest repo tag. Similarly, there is a chicken-and-egg condition for PR checking on contributions to MLOpsPython with updating the mcr.microsoft.com/mlops/python base image.

Preparing a PR to propose a solution to the above.

[dtzar]

I believe this is covered now.