Merge subarrays

[zonca]

• [x] notebook to produce test dataset
• [x] merging pipeline step
• [ ] level 3 pipeline that merges back subarrays into full frames

[zonca]

@ikashell @arunsurya77 I have a first version of the merge_subarrays functionality. For now this is trivial as it only puts back the array data, dq, and err slices back into place.

Do you have a description of the detailed algorithm we want to implement with the correct weighting for each component?

[zonca]

see an example on how it works now: https://gist.github.com/zonca/780d746964345776b91f5f6a4b558224

[arunsurya77]

This is great!. By weighting do you mean for the subarrays? I am not sure if there needs to be weighting for subarrays since its finally reflects the counts and the total exposure times remain the same for all the subarrays.

[zonca]

@arunsurya77 I was wondering if we need to scale the err array of the subarray, but now that I think more about it, probably not. whatever is in the reduced subarray frame is already the best estimate of the error and shouldn't be modified here.

[zonca]

ok, unit tests are passing: https://travis-ci.org/github/oirlab/iris_pipeline/builds/676362126

[arunsurya77]

We don't need to scale the err array. I think the merging will just keep individual pixel errors.

[zonca]

confirm we don't need to weight them differently