Improve (or remove) internal "registries"

[braingram]

Follow up from: #656

The "registries" inside stnode are difficult to use due to:

• the population of registries occurs in a separate file from where the registries are defined making it difficult to trace and coordinate how a registry is populated
• their mutability has led to previous issues where unintentional changes were had to track down

They can be removed with minor code restructuring as seen in the linked closed PR but will touch code that is likely to be changed while addressing:

• Sunset schemas rad#860
• FPS/TVAC datamodels being auto converted into a ScienceRaw model rad#869
• Do we benefit from dynamic node creation? #655