Drivers -- separate data "storage scheme" from "results"

[ceholden]

If TSTools hopes to be flexible with data "storage schemes"[1] and wants to be able to run these data against many time series algorithms, then it should do a better job isolating the data reading from the time series model running.

Related to #63, but provides a (better?) alternative plan. When a user configures TSTools against a dataset, only specify what kind of storage scheme the dataset uses and add this dataset as a "series". Separate from this data scheme configuration, allow the user to select the time series algorithm used.

Redesign configuration GUI to enable:

1. Select storage scheme type (stacked, AGDC, etc.)
   * Parameterize this storage scheme against a dataset
2. Allow user to add additional storage schemes (e.g., to add in met data, ALOS data, etc)
   * These additional storage schemes will appear as additional "Series" in TSTools
3. After data is configured, allow user to specify the time series algorithm to run against this data

On the backend, we'll need to ditch the concept of a time series "driver" as containing these two sets of concepts. In the long run it should make the pairing more explicit and will cut down on the overall number of "drivers".

[1] A "storage scheme" refers to how time series of spatial data are stored. For example, a "stacked dataset" storage scheme is a collection of multi-band images on disk with uniform geotransform/projections, each observation in the time series in a separate image. This "stacked" scheme is currently the only supported scheme.

[ceholden]

Mock up:

(add a section at bottom of left pane for TS algorithm selection)

See QgsCollapsibleGroupBox class for folding/unfolding of "Series" configuration sections