fix(microflows): preserve ELSIF arms by lowering to nested IfStmt (#745)

[mlehane1]

Fixes #745.

What

buildIfStatement (mdl/visitor/visitor_microflow_statements.go) read only exprs[0]/bodies[0] and the trailing ELSE body, so every middle (ELSIF expression THEN microflowBody)* pair from the grammar was silently dropped from the written model — check and exec reported success and mx check was clean, making the data loss invisible.

Mendix has no native elsif construct, so this lowers each ELSIF arm into a nested IfStmt in the ELSE branch of the arm before it (built innermost-first) — the canonical nested-exclusive-split shape Studio Pro users build by hand. Plain IF/ELSE builds exactly as before (regression-guarded).

Validation

• Unit tests (new mdl/visitor/visitor_microflow_elsif_test.go): elsif chain preserved with conditions/bodies at each level; elsif-without-else (innermost arm has no ELSE); plain if/else unchanged (no gratuitous nesting). Written failing-first against the old visitor.
• Bug-test script: mdl-examples/bug-tests/745-elsif-arms-dropped.mdl — exec against a fresh mxcli new 11.11.0 app, then DESCRIBE MICROFLOW BugTestElsif.MF_Grade shows all four arms (one/two/three/other) as a nested if/else chain with sensible auto-layout positions. mx check = 0 errors on the resulting app (Windows, mx 11.11.0).
• Suites: go test ./mdl/... ./sdk/... green; go vet ./mdl/... clean (grammar regenerated with ANTLR 4.13.1).
• Agentic testing: found, root-caused, fixed, and end-to-end verified via Claude Code driving mxcli headless (exec → describe round-trip → mx check) against real .mpr apps.
• Symptom row added to .claude/skills/fix-issue.md per the bug-fix checklist.

Opened as a draft pending maintainer sign-off on #745 per CONTRIBUTING.md.

🤖 Generated with Claude Code

[ako]

Approve — solid, well-scoped fix

Reviewed the diff against the codebase and ran it end-to-end. This correctly fixes the silent ELSIF-arm drop (#745).

Correctness

• The approach is the only correct one. ast.IfStmt (ast_microflow.go:289) has only Condition/ThenBody/ElseBody/HasElse — no ELSIF field — and Mendix has no native elsif. Lowering each arm into a nested IfStmt in the preceding ELSE is the canonical nested-exclusive-split shape and the only representation this AST supports.
• Downstream already handles it. A nested IfStmt inside an ElseBody is exactly what a hand-written nested if/else produces — bodyTerminates, layout.measureStatements, and the flow builder all recurse generically. No serializer/executor changes needed, correctly omitted.
• Edge cases traced: plain-IF, IF/ELSE, IF/ELSIF (no else), and IF/ELSIF/ELSE all lower correctly through the innermost-first loop; the plain path produces no gratuitous nesting; no annotation regression (caller applies s.Annotations to the returned outer stmt as before).

Testing (ran locally)

• New visitor_microflow_elsif_test.go (chain / no-else / plain-if regression), written failing-first: pass. Full ./mdl/visitor/ and ./mdl/executor/: pass. Build + vet: clean.
• End-to-end round-trip on a real 11.12 .mpr, both engines: exec + DESCRIBE MICROFLOW returns all four arms (one/two/three/other) as nested if/else on modelsdk and legacy. Before the fix only one/other survived.

Checklist

Test-first ✓, bug-test mdl-examples/bug-tests/745-… (numeric-prefix convention matches) ✓, fix-issue.md symptom row ✓, atomic single-concern diff ✓.

Minor, non-blocking

1. DESCRIBE renders the chain as nested if/else, not the elsif keyword — a faithful-but-lossy round-trip. Inherent (Mendix has no elsif) and accepted in the issue; worth a one-line note in the PR body so users aren't surprised the elsif source doesn't come back verbatim.
2. hasElse := … || ifCtx.ELSE() != nil is belt-and-suspenders (an ELSE always yields an extra body node) — harmless, mirrors the old fallback.

Nothing blocking. Merge once un-drafted per the #745 sign-off note.