Fits \(\log error = c + \gamma \log n\) and compares gamma with
the theoretical exponent. Rate theorems have no single-sample
value, so the measured exponent makes the claim checkable rather
than asserted. Mirrors
morie.fn._horowitz.check_rate.
Examples
n <- c(100, 400, 1600, 6400)
morie_rate_check(3 * n^-0.4, n, -0.4)$observed_exponent
#> [1] -0.4
