correctness: make autoFIPC itemtype-vector control flow length-safe

[seonghobae]

Finding

Protected master@f87c2324f1686135e57d8730c1b0b9420874f300 explicitly accepts itemtype with length 1 or nItems, but autoFIPC() later feeds vector comparisons into scalar control flow. Examples include:

if (itemtype == '3PL' && length(oldformBILOGprior) == 0) { ... }
...
if (itemtype == 'Rasch') {
  NewScaleParms[NewScaleParms$name == "a1", "est"] <- FALSE
  OldScaleParms[OldScaleParms$name == "a1", "est"] <- FALSE
}

R-devel documents if(cond) as requiring a length-one logical and treats longer conditions as an error; &&/|| likewise error when operands have more than one element. This means the public validation admits an nItems vector that later fails in ordinary control flow before the intended per-item mirt semantics can be honored.

Authoritative reference: R Project, base Control / Logic documentation: https://stat.ethz.ch/R-manual/R-devel/library/base/html/Control.html and https://stat.ethz.ch/R-manual/R-patched/library/base/html/Logic.html

RED

Add a realistic regression using a 2+ item form where itemtype has exactly nItems entries. At minimum cover:

• homogeneous vector, e.g. rep("2PL", nItems), proving the accepted vector form no longer dies at the first 3PL/control-flow predicate;
• mixed vector containing Rasch and non-Rasch items, proving the form-wide Rasch branch does not freeze every a1 simply because one item is Rasch;
• mixed ideal/non-ideal vector, preserving convergence checks whenever any non-ideal item requires them;
• mixed 3PL/non-3PL behavior with BILOG-prior handling defined explicitly rather than by first-element coercion.

Do not weaken the current input-length validation and do not replace the defect with itemtype[1].

Repair acceptance

Use explicit form-wide/per-item predicates whose semantics are testable (all(...), any(...), or a named helper) instead of passing a vector into if, &&, or ||. Preserve mirt::mod2values() item-level structural est flags for mixed item types; any additional global override must only apply when its invariant is truly form-wide. Add regression coverage before changing production code, then require exact-head R CMD check / owned tests / security gates and independent review.

No force rewrite, no warning suppression, no test fixture that reduces itemtype back to length 1.

@jules Please take this as a bounded TDD correctness repair from current protected master. Keep #349 and other active PR semantics separate; do not copy their unrelated deltas into this branch.

[seonghobae]

@jules Bounded owner handoff remains actionable from protected master@f87c2324f1686135e57d8730c1b0b9420874f300; no repair PR or owner response is visible yet. Please start with the RED before production edits: exercise an itemtype vector of length nItems across homogeneous 2PL and mixed Rasch/non-Rasch, ideal/non-ideal, and 3PL/non-3PL cases. Repair scalar control-flow predicates without itemtype[1], preserve item-level mirt::mod2values() structural flags, and keep #349 deltas out of this branch. Exact-head R CMD check and independent review are acceptance, not predecessor evidence.