Make cohorts more workflow-engine and cloud storage friendly

[armish]

1. Implement helper functions/wrappers to allow working against a bucket instead of an `NFS *server. Mostly addressed by Google Storage file access add-on for cohorts (a.k.a. minibucket) #219.
2. Allow more flexible Sample, Patient, and Cohorts object creation procedures (c.f. motivation: Annoying conflicts between patient-attributes & variables in additional_data #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 Multi-sample API #211 and Support easy Cohort creation #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).
   ...