Generalize from hardcoded 2nd-order to arbitrary-order derivatives via
recursive nesting. New API: dual_variable_n(), dual_constant_n(),
deriv_n(), differentiate_n(). Remove Rcpp/C++ fast paths — package is
now pure R, aligning with its unique value proposition of exact
machine-precision derivatives at any order (not speed).

- Add dual_variable_n(x, order) / dual_constant_n(x, order) for
  recursive nesting to arbitrary depth
- Add deriv_n(d, k) to extract k-th derivative from nested result
- Add differentiate_n(f, x, order) convenience evaluator
- Deprecate dual2_variable, dual2_constant, value2, first_deriv,
  second_deriv, differentiate2 as thin wrappers
- Delete src/, R/RcppExports.R, tests/bench/ — no compiled code
- Remove .is_scalar_dual() and all C++ branch gates
- 141 higher-order tests (3rd/4th/5th/6th order, multi-parameter
  non-linear functions, compositions vs numerical FD)
- 787 total tests pass, R CMD check: Status OK

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>