Validate every type set reference in EMV2 type products 🤖

[lwrage]

Fixes #2743

EMV2 type products only reported a type set when it was the first element, and the error highlighted the entire product. Check every element and attach each diagnostic to the indexed type reference, retaining the message that identifies the offending type set.

Issue2743Test uses external AADL fixtures to cover first, middle, and last positions, multiple type sets, and a qualified alias. It checks all seven expected errors, their messages, and the exact highlighted reference text. Separate valid cases cover error-type products, single types, standalone type sets, aliases, and unions of type sets. Before the fix, the invalid fixture produced only two errors instead of seven; the valid cases passed.

Validation:

• Focused regression: 2 tests passed.

  mvn -o -T5 -s releng/osate.releng/settings.xml -Plocal \
    -pl :org.osate.xtext.aadl2.errormodel,:org.osate.aadl2.errormodel.tests,:org.osate.xtext.aadl2.errormodel.feature \
    -Dtycho.localArtifacts=default -Dpr.build=true -Dsign=false \
    -Dspotbugs=false -Dcodecoverage=false -Djavadoc=false \
    -Dtest=Issue2743Test -DfailIfNoTests=false clean install

• Full EMV2 suite: the same command without -Dtest=Issue2743Test passed 295 tests.
• Clean root reactor: all 143 projects succeeded; Surefire reports contain 1,670 tests across 15 bundles, with zero failures, errors, or skips.

  mvn -o -T5 -s releng/osate.releng/settings.xml -Plocal \
    -Dtycho.localArtifacts=ignore -Dpr.build=true -Dsign=false \
    -Dspotbugs=false -Dcodecoverage=false -Djavadoc=false \
    -DfailIfNoTests=false clean install

• git diff --check passed; the worktree is clean.

The regression commit precedes the production fix. This PR targets master and has no dependencies on unmerged PRs. Risk is limited to validation: previously missed invalid references now receive errors; grammar and model representation are unchanged.