use proportional K600_daily_sigma_frac

[aappling-usgs]

For Kb and Kl models, we currently have

K600_daily ~ normal(exp(K600_daily_predlog), K600_daily_sigma)

I'd now like to make K600_daily_sigma a proportion of the K600 prediction on each day:

K600_daily_sigma = K600_daily_sigma_frac*exp(K600_daily_predlog)
K600_daily ~ normal(exp(K600_daily_predlog), K600_daily_sigma)

where in this version the user will specify the value of K600_daily_sigma_frac rather than K600_daily_sigma. In a sense, this is a cross between a normal and a lognormal, because there can be bigger deviations for larger predlog values, but they won't be too too big.

[aappling-usgs]

Default of something like 5% would be good.

Charles does not endorse this approach, because high-flow days within a site should probably be even more similar to one another than low-flow days are (assuming variability on low-flow days is driven by things like macrophytes and litterfall). He endorses sticking to K600_daily_sigma (non-log, non-proportional, good ol' normal distribution).

[aappling-usgs]

Don't do this. But do see #312 for idea on how to help user tailor the value to their site[s].