gadget3 0.10-0:

New features

Add g3_parameterized('x', by_step = TRUE) for seasonal parameters #115
Support R factor levels as well as MFDB-style attributes for likelihood observation data. This means dplyr group_by(length = cut(...)) can be used as well as MFDB for groupings #112
Parameterised defaults for g3l_understocking(), g3a_naturalmortality(), g3a_renewal_normalparam()
Add g3a_renewal_vonb_t0() & g3a_renewal_vonb_recl(), with g3a_renewal_vonb() being an alias to the latter

Remove scale = 0.001 from default K parameters in g3a_grow, g3a_renew_*
g3a_renewal_vonb_t0() is now the default mean_f for *_normalparam()
g3a_initialconditions_normalparam() now offsets any age in mean_f (i.e. the VonB formula) by cur_step_size, in effect running at step -1.

g3_suitability_andersen() now produces sensible values - https://github.com/gadget-framework/gadget3/issues/108
g3a_age() now supports stocks with a single age (i.e. minage == maxage)

Add recage parameter to g3a_renewal_vonb / g3a_renewal_initabund
Add g3_suitability_andersenfleet(), a fleet-specialised andsersen suitability function.
Make g3_is_stock() public
g3_eval(), to evaluate snippets of a gadget3 model.

g3l_distribution_*(transform_fs = ...) now happens before aggregation, not after. Any matrix used now has to be expressed in terms of the stock, not aggregated age.
g3_param_table() now returns NaN (and warns) on a missing value, instead of aborting.
optim(g3_tmb_par(...)) is now an error. When optimising, always use obj.fun$par.

Support as.vector(array) in TMB formulas, so arrays can be used with TMB vectorized functions, e.g. pnorm(as.vector(ar[,1])).