[v2] simpeg inversion stability/repeatability problems due to libaarhus precission

[mp-emrld]

EM Data which are processed with the UX/UI and using a notebook and then stored on disk differ slightly, see also this notebook cells 19-24

That causes (simpeg/simpleem3) inversion of these two datasets to diverge and the result differs significantly:

• Inversion of notebook processed data
• inversion of UX processed data

However, when rounding the input data sets prior to inversion the inversion results are equal:

• inversion of notebook processed and rounded data
• inversion of UX processed and rounded data

A possible Solution would be

• use scientiffic notation with max 5 digits for the data Gate_Ch01 , Gate_Ch02 and the STD's STD_Ch01, STD_Ch02 stored in the xyz files on disk
• same notation precision is used for the official data delivery from SkyTEM (see example here)

[redhog]

Better solution is to save intermediate data as binary (msgpack?) so no rounding occur.