Support multiple python virtual environment in a CM workflow

[anandhu-eng]

---

Migrated from mlcommons/mlperf-automations_archived#31
Originally created by @arjunsuresh on Fri, 15 Mar 2024 14:04:59 GMT

---

Sometimes in a single workflow we might need to use a different python environment. A use case is for Intel MLPerf inference implementation where calibration step needs a different python environment. Currently this is tricky in CM workflow as we use --adr to update all python dependencies when we detect an environment.