add milestone test

[LittleBeannie]

Closes #197

[nanxstats]

I replaced your ifelse() calls with vanilla if-else constructs:

• In non-vectorized cases, ifelse() is often much slower than if-else because ifelse() simply does too many things. In this particular formula, if-else is 4x faster than ifelse().
• You can also write such formulas in a concise way by upcasting the logical variables, although it is a tad bit slower than if-else in this case.

library(microbenchmark)

f_ifelse <- function(fit_res, na_col, na_exp) {
  ifelse(na_col, 0, fit_res$std.err[1])^2 + ifelse(na_exp, 0, fit_res$std.err[2])^2
}

f_if_else <- function(fit_res, na_col, na_exp) {
  term1 <- if (na_col) 0 else fit_res$std.err[1]
  term2 <- if (na_exp) 0 else fit_res$std.err[2]
  term1^2 + term2^2
}

f_upcast <- function(fit_res, na_col, na_exp) {
  ((!na_col) * fit_res$std.err[1])^2 + ((!na_exp) * fit_res$std.err[2])^2
}

set.seed(42)
fit_res <- list(std.err = rnorm(2))
na_col <- FALSE
na_exp <- TRUE

f_ifelse(fit_res, na_col, na_exp)
f_if_else(fit_res, na_col, na_exp)
f_upcast(fit_res, na_col, na_exp)

microbenchmark(
  `ifelse` = f_ifelse(fit_res, na_col, na_exp),
  `if-else` = f_if_else(fit_res, na_col, na_exp),
  `upcast` = f_upcast(fit_res, na_col, na_exp),
  times = 5e+5
)

#> Unit: nanoseconds
#>     expr  min   lq      mean median   uq     max neval cld
#>   ifelse 1312 1435 1715.1805   1476 1640 7033919 1e+05   a
#>  if-else  287  369  424.8617    369  369 1164113 1e+05   b
#>   upcast  328  410  493.1685    410  451 2044342 1e+05   b