n-dimensional integrations in models

[levisc8]

@wpetry had mentioned this before, and came up today in discussion w/ Tomos. Currently, the par_set_indices method won't work, because each level of sex will be treated as a separate population when population size/lambda is computed.

Likely requires an additional argument to init_ipm() (e.g. init_ipm(..., two_sex = TRUE), and then altering .init_iteration and .pop_size to be generic on that. Almost certainly overlooking a critical component though.

[levisc8]

name change to reflect new approach.

functions that will need updating:

• define_pop_state()
• define_domains()
• .pop_size()
• .init_iteration()
• .find_start_end()
• .fun_to_iteration_mat()

This one needs a deep rethink and probably a full blown replacement. state_var_start/state_var_end now need to be vectors of arbitrary size. Probably best to change to dim(out) <- dims, where dims <- c(n_mesh_1, n_mesh_2, ..., n_mesh_x), generating arrays of arbitrary dimension.

functions that need creation:

• .create_iter_exprs.n_dim_int() (probably name this class something else)
• .combine_iter_exprs.n_dim_int()
• .make_iter_right/left.n_dim_int() (to work with left/right_ev()

to investigate further:

• .iterate_model.*

DETERMINISTIC METHODS: These may need updating because of the way that parameter set indexed models are handled. I don't remember exactly how the pop_state/sub_kernel names are going to look when they reach this point, so may need to add a line or two to accommodate 2-sex models.

• right_mult()/left_mult()

These will need generalization to higher dimensional objects. see literature on n-dimensional array multiplication.

To be modified further...