rdmat extra calculation 'issue' or preference for byte offset formats

[cx1111]

Aside from creating a map of nans for different formats, I changed the way byte offset format signals are converted into physical units.

Old: baseline=baseline+constant so when doing val=(val-baseline)/gain, you're actually subtracting the constant and getting the right value.

New chosen:

• Do not add this constant to baseline variable.
• Subtract the constant from the variables val=val-constant
• Convert the nan mapped values to nans.
• Do val=(val-baseline)/gain.

This gives an extra subtraction calculation/computation but is technically correct.

New alternate option:

• Add constant to baseline variable
• Convert (nan mapped values + constant) into nan
• Do val=(val-baseline)/gain

This is faster but the key-value pairs for wfdbnan will be technically incorrect for format 80 and 160 and may confuse people?