The choice of using np.float16

[rogerkuou]

Should we use np.float16? If yes, to which variables?

This should happend in the step where we create Zarr SLC stack. See script here

The motivation is to save disk space.

Current status (updated on Oct 21, 2024)

• real and imag fields for complex data:
  • Currently their are saved as separated fields, both in np.float16
  • ToDo: eveluate if they provide sufficient precision
• h2ph values: at present we conclude the precision is not enough. See discussin in this PR.