Propagate call for error messages

[teunbrand]

This PR aims to fix #379.

Briefly, it adds a call argument to range training that is used to communicate error messages.
Short demo:

devtools::load_all("~/packages/scales")
#> ℹ Loading scales

i_train_ranges <- function(x) {
  range <- ContinuousRange$new()
  range$train(x, call = current_call())
  range$range
}

i_train_ranges(1:10)
#> [1]  1 10
i_train_ranges(c("foo", "bar"))
#> Error in `i_train_ranges()`:
#> ! Discrete value supplied to a continuous scale

^{Created on 2023-12-22 with reprex v2.0.2}

[thomasp85]

LGTM