Resolve pip v conda in tests

Conda has been a helpful tool for managing dependencies in test environments, especially given its ability to specify python versions. However, it has become a burden, specifically relating to:
1) Issues with OpenMind downtimes, conda breaking often
2) Running pip install way too often during testing, slowing down process and introducing bugs such as "installing UNKNOWN package"

Goal:
- Remove conda from travis and jenkins tests, replace with venv or similar
- Identify workarounds for python version specification where needed