abstractions and implementation for remote dataset support

[stuckyb]

Currently, dataset abstractions (e.g., dataset abstract base classes) are geared toward local datasets. A similar framework needs to be designed and implemented for remote datasets and unified with the local dataset abstractions. Thus, this issue is both a major software engineering task and implementation task.

[HeatherSavoy-USDA]

Will we have pre-populated catalog entries for remote datasets like the local ones? Or generate/search metadata as needed? If the former, we could at least have data provider- and/or data access protocol-specific ingestion processes (#34)?

[HeatherSavoy-USDA]

I've been exploring how to support the remote MODIS NDVI dataset (#24). The roadblocks I've run into have been mentioned in various issues but I'm listed them here:

1. [Minor] It would be nice to have a general self.ds_path that supports remote data URLs (mentioned in configuration system #44)
2. Currently getData() is opening the remote dataset for every time increment, but it would be more efficient to point to the data store once (stacking instead of explicitly looping over layers in the same dataset? #31)
3. The data are sparse daily (sub-monthly but non-daily datasets #35) so requests on a specific date of known data is currently working, but a range of dates is not due to Roadblock 2.
4. I'm doing a slice in the x/y dimensions based on the bounds of subset_geom.geom to limit the amount of data downloaded. It would be better/safer if the geometry buffer was implemented (reproject clipping polygons to dataset CRS #38).

[HeatherSavoy-USDA]

I think the only thing remote data specific left that I've run into with the MODIS NDVI dataset that hasn't been resolved as of now is self.ds_path not playing nice with URLs pointing to data stores. I'm currently just not using it so it's not a roadblock, but I'd like it to work to make the code looks nice. I can pass a URL into the initializer like

super().__init__('https://thredds.daac.ornl.gov/thredds/dodsC/ornldaac', '1299')

But it strips the '//' out and doesn't play nice with open_url(). It seems like a simple fix that has different options on how to do, but it seems like a python specific style choice that I'll leave to a more python person.