[ENH] post/get datasets

[jdkent]

ref #28
@tyarkoni: should the nimads_data dictionary be nested and contain all the analysis information, images, points information or should the dictionary merely be a pointer to a list of study ids?

If the nimads_data dataset is nested then the information should be self-contained in the dataset, but all the data will essentially be copied over, meaning if the data is changed on the database then the changes will not be propagated to the nimads_data dictionary.
e.g.

{
    "dataset": [{"analysis": [...], "name": ..., ...}, {"analysis": [...], "name": ..., ...}]
}

If the dataset is just a list of study ID's, then the dataset may change depending on updates to the database, and the client library will handle resolving the study ID's into their actual representations.
e.g.

{
    "dataset": ["sldkfnejfne", "erfewrgrtrt", "apofiermkgt"]
}

It sounds like I want the first option since we want datasets to be immutable (and that's what presented in #28), but I could see the motivation for just a list of Study IDs since it's more compact representation.

[tyarkoni]

Yes, we want #1. The goal here is self-containment and reproducibility. Redundancy is fine, as is the existence of multiple out-of-sync versions of the same study in different datasets.