Allow the vs record's precision to be specified by the PREC field

[kmpeters]

Allow the vs record's precision to be specified by the PREC field.

Commit eef412e includes changes suggested by Claude Opus 4.6 via Argo.

The changes "work" and have been tested at 3ID-B, however, it isn't obvious to me if other modifiations are required. I'm not familiar enough with all the devices supported by the vs record, so I don't know which fields in get_precision should be controlled by the PREC field and which, if any, should remain hard-coded.

[anjohnson]

FYI: The basic record types in Base use PREC for VAL and related fields (OVAL, PVAL etc.), but call recGblGetPrec(paddr, precision) for almost everything else. That gives 15 for fields that are floating-point, and 0 for the rest (those probably should be 7 for float and 15 for double, but that isn't what's currently implemented, and we don't have any DBF_FLOAT fields in the Base records anyway).

Some record types have a global variable which can be set with the iocsh var command for the precision of specific fields such as boHIGHprecision for the HIGH field of the bo record (a double holding the pulse width in seconds) — I didn't want to make those a fixed number, and this lets individual IOCs override the default of 2.

Other records such as the aSub, calc and sub that have INP links ask the link what the precision is of the link's target to provide the precision for the local field fetched from that link. They have similar code for returning unit strings too, which is pretty cool if you have a display manager which actually shows the unit strings properly.

Then there's the mbb[io]Direct records, which don't have a PREC field or provide a floating-point value, but they return the value of the NOBT field for VAL. Until I ran git grep just now I didn't know about that at all. I doubt if anything actually uses that, although it's possible that something like the MEDM Byte widet might.