Fiber attenuation: total loss vs. loss-per-km

[jktjkt]

Today, fiber loss is specified in the IO format indirectly, via the length and the loss_coef. Vinay raised a good point in that users typically have access to OTDR measurements, and these report a total loss value. Should we switch to working with the total loss of the fiber (except the connectors) in the YANG interface? That way, people won't have to divide that number from the OTDR measurement by the fiber length.

The format currently supports specifying the loss either as a scalar value (which applies over the whole spectrum), or as a mapping of frequency→ loss_coeff; this will have to be preserved either way.

Cc: @sme791