Openmsi upload

[oruebel]

I added two functions to the ArrayedImage class.

• saveToOpenMSIFile(...) allows saving of ArrayedImage data to OpenMSI compliant HDF5 file
• upload_to_openmsi(...) allows uploading ArrayedImage data as an OpenMSI compliant HDF5 file back to NERSC and adding the file to the OpenMSI website.

What does this allow us to do:

• Share results via the OpenMSI website
• The files will appear as usual in the file browser on the website
• We can retrieve data from the files via the REST API using qmetadata and qcube

I think this is a pretty nice feature that at least from the paper perspective allow us to close the loop from getting data from OpenMSI to sharing the results again via OpenMSI.

What is not yet supported:

• While the name of the input file etc. are stored, they are currently not saved as dependencies so that the website won't show a link to the file
• The results are currently stored in a separate HDF5 file that is independent of the input file. Adding results generated outside of NERSC back into files that are managed by OpenMSI will require more effort as well as changes to OpenMSI itself. To make what we have here work, I did not have to change anything in OpenMSI itself.
• There are no viewer functions for this yet so you won't be able to use the online viewer (only the file browser and manager) with OMAAT results
• It would be cool to also save the JSON (and HTML) of the full notebook in the file. This shouldn't be a problem, but I need to figure out how we can get the full document from iPython or nbconvert. I think this should also allow one to then directly look at the notebook in nbviewer.
• The more advanced provenance and reexecution features of BASTet are not yet supported for OMAAT. For this I would need to extend BASTet to be able to wrap and run notebooks directly.

Python 2/3:
The tool itself should still work in both Python 2 and 3, however, these two functions will currently only work in Python 2 because they need BASTet (https://github.com/biorack/BASTet) which hasn't been ported yet to Python 3.