unitSI: relax type?

[ax3l]

Short Description

@DavidSagan proposed in marcguetg/h5particle#11 to remove the constrain of double precision for the unitSI attribute.

This increases flexibility a bit and increases reader-logic as discussed in the link above (we can summarize this here in an update of the PR description). In principle, the precision of the data type that is described is high enough for this attribute (e.g. single precision) and some applications might actually want to use even for the conversion attribute a higher precision.

One could also think about defining that unitSI needs to be at least the same precision as the record (component) it describes.

Side Effects with Other Proposals

The progress on this proposal need to be synchronized with the development of the proposal in #155.

[ax3l]

ccing @C0nsultant: it's not high priority right now, do any opinion on the complexity of this?