Migrate publication processing and DB writing tasks to `models`

[brycefrank]

Large portions of allometric are dedicated to processing publications and, as we develop the database, handling conversion of model objects into JSON. This is making the "root" package quite expansive and muddying documentation for typical end users.

Instead, we can move these pieces of allometric directly to models, which is already an R package by definition. We need to ensure that moving publication processing, especially map_publications is feasible first. I will start this work on migrate_processing branches for each of the repos.

In general we should have the philosophy: models can depend on allometric but not vice versa, because most users will not want/need to install models directly.

[brycefrank]

Here is a preliminary list of things to consider

R

• load.R
  • all functions except combine_taxa, aggregate_taxa, load_models
• publication_processing.R
  • all can go
• boilerplate.R
  • all can go

testthat

• test-install.R
  • Keeping this in allometric. After all, users install models via allometric
• test-publication-processing.R
• test-update-models.R