Prepare for go

[omad]

Lots of changes in preparation for large scale runs:

• Includes all changes from #25 for improving the work generation step
• Detect and fail early if not using full resources in an MPI job
• Use MPI in a simpler manner. Split the input file by the number of workers, each MPI worker completes every n^th task.
• Use structlog for more orderly logging
• Detect whether running interactively to decide which logging format to use (Needs improvement for within PBS jobs, wrongly detects it as interactive)
• Allow skipping pyramids for some bands (TODO: breaks verify step)
• Clean up code for searching and verifying output COGs