Make cohorts more workflow-engine and cloud storage friendly

[armish]

1. [x] Implement helper functions/wrappers to allow working against a bucket instead of an `NFS *server. Mostly addressed by #219.
2. [ ] Allow more flexible Sample, Patient, and Cohorts object creation procedures (c.f. motivation: #220) and try to extract data reading/manipulation logic out of the main cohorts module to keep it as much clutter-free as possible (related #211 and #218).
3. [ ] Provide some basic sanity check scripts/solutions out of the pip to prevent common/known pitfalls to be captured upon common git commands (will demo potential use cases for this soon). ...