Simplify Dataset Class

[jadball]

The dataset class/file has, at a minimum:

• Data
  • Motor positions
• File paths
  • dataroot
  • analysisroot
  • sample name
  • dataset name

Right now, we include a bunch of other stuff:

• File paths to processed data
  • Messy and not very phase-aware, don't like this at all right now
• spatial files
• parameter files
  • pointer to json
    • itself a pointer to .par files

Additionally the H5 file is not human-readable as text...
Instead I propose this:

• Dataset file
  • This becomes ONLY the data needed to construct a sinogram (dty/omega positions)
• Single Analysis Schema / Project Schema / Whatever file
  • See Extended Analysis Schema #372

This can be partially autogenerated during the acquisition stage or when the first notebook (/ewoks) is run. Detector parameters filled in on the BLISS side, etc

The file paths issue is then solved by keeping the notebooks in the same folder as the files we create, so relative paths only are needed
If you make a peaks1.h5 and a peaks2.h5, you just need to update the indexing notebook to say which peaks file you want to use
This means we can copy the entire analysispath to different locations and nothing breaks.

[jadball]

Tasks needed:

• Remove file management stuff from the dataset class
• Update notebooks accordingly

[jadball]

Should wait until #371 is complete before we do this.