Julia BRT improvements

[frousseu]

The main issues right now with the Julia BRT pipeline are:

1 - It can only work with the 4326 epsg 2 - You cannot choose the species and it can only be run for a pre-selected species 3 - It does not use a target-group for taking into account the sampling bias

@tpoisot, I think your input could be mostly for the first point, since it appears to be Julia related. For the third point, the Julia BRT pipeline could be connected with the selectBackground.R script which has the option for a target group using the gbif density rasters or the option could be added in the generateBackground.jl. Otherwise, I can look into 2- and 3- if we are not adding the target group in the Julia script.

[tpoisot]

1. It currently works with any CRS but falls back to EPSG:4326 internally, but I am almost done "fixing" this
2. sounds like a simple fix @gottacatchenall can do in the pipeline code?
3. there's a function to do background sampling by distance to occupied cells (after spatial thinning) we can use, if you would like an alternative method and there's broad consensus on it, send a reference and I'll write it up

[frousseu]

The general idea for the target group approach https://doi.org/10.1111/j.0906-7590.2008.5203.x is to sample from a large group of species with similar sampling bias to generate the background points. In our case, we have the gbif occurrence density rasters for general groups (plants, birds, etc. https://coleo.biodiversite-quebec.ca/apps/io-layers/gbif_heatmaps/all-heatmap) from which we can sample to generate background points.

[tpoisot]

I see - I'm not 100% convinced the statistical assumptions of this specific technique match with non-MaxEnt methods, but I'll read more. Anyways if this is just about using a layer of densities, this is already supported by the package and would be extremely fast.