Veg models predict in the S when species is not there

[psolymos]

• check all sector, boot, and clim data for consistency
• subset for modeling region only: this will speed up loading as well

[psolymos]

Possible solution, that relates to some other issues with multi-species aggregate results:

• keep separate results for each taxa (implicitly identified by the set of species): this is needed because overall intactness is an average of each taxa (weighted as 1 and not by # species in taxa); this applies to SI and richness as well -- at least have the option of by_taxa=TRUE or something.
• exclude species from overall intactness as we do it for BMF at the subregional scale. This would require NR and NSR to be part of the .c4if environment when load_common_data, and would fall back to limit for the assessment.

[psolymos]

Using the new functionality from #24 we can do the following:

1. calculate SI
2. assess abund < limit threshold at the NSRxLUF level in the whole province and record in a NSRxLUF x SPP matrix as 0/1
3. make a 0/1 raster and multiply it with SI (it will get 0 where species does not count)
4. increment final raster with new species values
5. at the end, use the matched 0/1 matrix rowSums as denominator for SI average

[psolymos]

The N/S non-prediction issue needs to be dealt with at early stages, providing NA instead of 0 in loaded species data, and masked in rasters as well.