Utilize Conda virtual environments

[YevheniiSemendiak]

Why

Feature request came from the Reseach team, namely from @johngull.
This will give us many advantages right now and in the future:

1. We could export all (not only python) dependencies via the single environment.yml file, many tools like MLFlow / Seldon support Conda environments -- this helps with integrations.
2. Users could use conda channels to install non-pip dependencies (for instance, extra version of Python, GPU libraries requiring system libs, etc).
3. Later we could enable our users to build container images by providing a single environment.yml file and base image version.

Also, in the base image we could utilize mamba, a backend replacement for the Conda implemented in C++. It also works with the channels, same commands and UX. Just different CLI utility (conda -> mamba).