Allow combinations of initial parameters to fit for the best AIC or BIC

likethis

```
init_pars <-
 tibble::tibble(
  Linf = rep(c(200, 300, 400), each = 3),
  K     = rep(c(0.5, 0.6, 0.7), by = 3),
  t0    = rep(0, by = 9)
 )

fit_explore(init_pars, "vb")
```

AIC or BIC choose the best one. Make sure the range of K does not exceed  too dramatically.

