bboutools extensions

[sebdalgarno]

Extensions to bboutools - integrating changes made in https://github.com/LandSciTech/bboutoolsMultiPop and other requested changes from ECCC

In order of commits:

• updated data and models to use CaribouYear instead of Year - bug or feature? Changing the meaning of Year #75
• allowed setting niters = 0 in bb_fit_survival/recruitment(), which bypasses MCMC sampling but otherwise outputs the same returned object (including model, empty samples array etc.) - extensions - allow fit_ function to return model and data without MCMC sampling #85
• added exported bb_predict_x_samples() equivalent to all bb_predict_x() functions to return samples instead of coefficient table. Added tests for each new samples function - extensions - ability to return samples from predict functions #83
• fixed ML fit error (laplace approximation functionality moved to new package nimbleQuad) - nimbleQuad package now handles laplace approximation #88
• implemented support for multiple populations in Bayesian models - extensions - Allow for analysis of groups of populations with shared interannual variation #77 (uses same model formulation as bboutoolsMultiPop, but specific implementation differs).

With multi-population implementation, some outputs will appear different and this change is applied across the board whether it is single or multi-population.

If a user has scripts running previous bboutools functions on single populations, these functions will still run without failure (i.e., plotting, coef, predictions, etc.). Only issues arise if customized code was developed that:

1. references index values of parameter estimates (i.e. now bAnnual[1,1] instead of bAnnual[1].
2. references data.frame estimate output by column number position (i.e., due to new PopulationName column in data.frame).