fix up sar

[ajrominger]

first, check that it's really right (e.g. using python macroeco):

macecoS <- rev(c(133, 116.41715386, 102.63632568, 90.13796807, 78.30328489, 
                 66.90263572, 55.90125381, 45.38753706, 35.5445402))
meterS <- meteSAR(S0 = 133, N0 = 49489, A0 = 160000, Amin = 625)$pred$S
plot(macecoS, meterS)
abline(0, 1)
## it's a little off for intermediate A

Then make empirical SAR look cleaner by taking mean of observed S at scale A (instead of all S)