Option to utilize R-based `make_state` during calibration

[stevencarlislewalker]

Should be three options:

1. Provide a single state that is used as the starting point for all parameter values
2. Compute the starting state using the R-based algorithm (no Jacobian, and rounding)
3. Compute the starting state using the TMB-based algorithm (Jacobian, and no rounding -- round not happening due to automatic differentiation subtleties)

• [x] Add do_make_state flag in TMB data structure
• [x] Coordinate state construction options so that R-side and c++-side are working properly together