CIDEVSTC-291 - Coverage Projection

[lukecampbell]

Adds the ability to reference external coverages.

Necessary for:

• Colocated instrument in data processing
• Geospatial information from platform onto instrument.

[lukecampbell]

@mmeisinger Please review

[lukecampbell]

I want to point out some MAJOR shortcomings of this initial pull request.

• Depends on the data being SORTED (np.searchsorted is used), we'll have to refine it to make a much slower algorithm for the unsorted datasets.
• Data points that exist outside the domain are extrapolated using the last two data points and a linear model.
• This is going to completely fall apart when we get to M193

[lukecampbell]

Background

1. Colocated L2 Data Products
2. Lat and Lon for AUV/Gliders

Some of the particulars of the use case for this got flushed out in a meeting a few weeks back:

• I think we'll need to include not only a reference to another coverage but a window.
  • reference_id
  • extents (that are potentially open)

Initially the parameter will not have any references and should return all NaNs for each time-step in the local domain (the domain of the coverage this parameter belongs to). Once a device is deployed, we'll know which instrument is in proximity and should be referenced.

I believe, that we'll add logic inside activate_deployment that will identify a CTD on the same platform and then identify any parameters within the dataset/coverage that are of this type (ExternalFunction) and then add a reference to that parameter and the window of the deployment.

The same can be said about latitude and longitude.

[caseybryant]

This pull request should be closed without merging. It was replaced by #194. #195 has been created to capture edge cases that haven't been handled yet.
@lukecampbell Do you want to document any edge cases missing in #195 and close this PR?

[lukecampbell]

Once I get into the integration more, I'll look to document missing edge cases but, I feel the core of it is there and will meet our needs. I definitely need to think through the big picture and how this plays into multiple deployments.