Support VAST release 3.5.0

VAST release 3.5.0 involves:

• Adding feature for barrier-SPDE approach;
• Changing density covariates to index by X_gctp to allow treating cohort effects via the spatially-varying coefficient interface
• Eliminating the previous, deprecated interface to seasonal effects to simply CPP code

Improved EOF analysis, overlap metrics, and shapefile usage

FishStatsUtils 2.6.0 matches VAST release 3.4.0 and these include several improvements including:

1. Improved options for Empirical Orthogonal Function analysis
2. New options for calculating relationships among variables including overlap e.g., Schoeners-D
3. Options to define extrapolation-area based on user-supplied shapefile.

Improving projection options...

... by removing dependency on package PBSmapping and instead using package sp for all projections. Change is fully backwards compatible and by default auto-detects the same UTM zone as used by PBSmapping. Also fixing small bugs in make_covariates(.) and load_example(.)

Adding example for expanding stomach-content data

thanks Arnaud Gruss for testing the method and putting together the example!

Update version number in DESCRIPTION...

... but otherwise identical to release 2.3.3

Fix bug in `fine_scale=TRUE`...

... which occurred when observations were only available in subset of the extrapolation-grid, where the INLA mesh would only cover the spatial domain of observations and random-fields were not active for other portions of the extrapolation-grid

Fixed another bug in `make_covariates(.)` ...

which failed with some column orders for input covariate_data

Fix bug in `make_covariates`

Fixes bug in make_covariates as pushed in release 2.3.0. This bug occurred when the user was creating covariates based on the new formula interface (which hopefully had not seen much use yet) and covariate_data were not ordered with respect to year. I have added a new integrated-test that checks intercept and slope estimates using the conventional lognormal delta-model without any random effects (i.e., no spatial or spatio-temporal terms) against two separate lognormal and logistic GLMs fitted using stats::glm(.), so results using the formula interface are more carefully maintained in future releases.

Improve plots and interface

Improving plots and interface by:

1. Using sp package to plot variables
2. Simplifying passing optional arguments to fit_model
3. Adding formula interface for density covariates.

Adding S3 objects

And other smaller updates