Add support for CONDA_PREFIX when training DeLFT

[de-code]

Based on the GitHub conversation:

I wonder if we could read the CONDA_PREFIX and pick up python from that location instead of just calling a generic python3 executable...

In this way we maintain the double compatibility:

1. environment activation + command launch and
2. environment setting + command launch

/cc @lfoppiano

[de-code]

Descripton for CONDA_PREFIX from the docs:

The path to the conda environment used to build the package, such as /path/to/conda/env. Useful to pass as the environment prefix parameter to various conda tools, usually labeled -p or --prefix.

I thought CONDA_PREFIX was set after a conda environment was activated. But it seems it is also set by users to specify which environment they want to use.

The regular Python's equivalent is VIRTUAL_ENV to some degree. I didn't find much documentation for it. I don't think people set it to indicate which virtual environment to use. Instead it will be passed in. There wouldn't be much harm in using it though. As that should be the same after activating the environment.

We could use PythonEnvironmentConfig.java - or only support CONDA_PREFIX for this purpose.