rdborrow 0.0.4.1

Breaking changes

bootstrap_ci_type no longer accepts "stud" in any method constructor. Studentized intervals require a variance estimate for each bootstrap replicate, which the estimators do not produce, so the option failed whenever it was used (#77).

Corrected the DOIs in the package references. The two DOIs cited in DESCRIPTION pointed to unrelated papers, and the open-label extension methods and vignettes cited an unrelated paper rather than Zhou X, Pang H, Drake C, Burger HU, Zhu J (2024). Reported by Herb Pang.
bootstrap_ci_type = "bca" no longer errors. boot.ci() re-invokes the bootstrap statistic through empinf() without the arguments passed to boot(), so those are now captured in a closure (#77).
bootstrap_ci_type = "norm" no longer returns NA confidence bounds. The normal component of boot::boot.ci() output has three columns rather than five, so the fixed index used to read the bounds ran past the end of it (#77).
did_ec_ipw() and did_ec_aipw() no longer return NA estimates when trt_formula is left at its default of NULL. The marginal randomization probability was a scalar, so subsetting the treatment weights by a length-N logical produced NA for all but the first subject (#71).

setup_bootstrap() has been removed. Bootstrap settings are now specified directly in method constructors (e.g., ec_ipw(bootstrap = 500, bootstrap_ci_type = "perc")). Calling setup_bootstrap() now raises an error with migration instructions.
setup_method_weighting(), setup_method_DID(), and setup_method_SCM() have been removed. Use ec_ipw(), ec_aipw(), did_ec_ipw(), did_ec_aipw(), did_ec_or(), or scm() instead. Calling the old constructors now raises an error with migration instructions.
EC_IPW_OPT(), EC_AIPW_OPT(), and all legacy_DID_EC_* / legacy_SCM* internal functions have been deleted.
did_ec_ipw() and did_ec_aipw(): trt_formula default changed from "" to NULL.

ec_ipw() and ec_aipw() QC’d against Zhou et al. (2024a, JRSS-A). Internal code now maps directly to paper equations (Def 1/2, Eq 6/7, Theorems 3/4).
All estimator internals refactored: redundant parameters removed, potential trick replaced with direct weighted means, consistent naming (_core, _se, _boot_statistic).
build_analysis_df() is now an S4 method on method_weighting_obj.
Bootstrap logic moved from .format_primary_results() (deleted) to inline control flow in each estimate() method.
Added input validation: S/A must be binary 0/1, T_cross must be a positive integer less than the number of outcomes, outcome_formula length must match number of outcomes.
Vignette improvements: T_cross semantics documented, head() calls added, rank-deficiency warnings suppressed in OLE simulation output.

New API introduced: eg ec_ipw() and ec_aipw() replace setup_method_weighting().
Internally, core functions such as EC_IPW_OPT() are further broken down into atomic functions.
Bootstrapping functions now call atomic functions (instead of similar, duplicated code).
Additional test cases in preparation of deprecating legacy_ functions such as EC_IPW_OPT().
Bug fixes (AIPW called the IPW bootstrapping function)