investigate yaml serialization

[yxqd]

In the long run, we should use hdf file and define a schema for saving intermediate and final results in the processing chain. At this moment, however, we may need a quick solution to at the least store intermediate results with enough meta data. One use case is the sinograms. It may be OK to save all sinograms in a series of images, each corresponds to one layer. However, how should we save the rotation angles? Ideally we should have axes in the hdf file storing such information. For now, maybe we can save that kind of information using a yaml file.

The counter argument would be that for these intermediate files, we should just define the hdf5 schema now.