[DanielThomas]

Summary

Fix correctness issues identified while reviewing Guillaume Laforge's formatter comparison, plus an additional import-binding defect found during reproduction.

• Respect explicit type imports during ordinary and static wildcard expansion. Previously, the CLI could change a field's type from java.lang.Thread to a wildcard-imported custom.Thread while leaving the result compilable.
• Retain used explicit java.lang imports instead of unconditionally removing them. Removing them can introduce ambiguity or change bindings; unused imports are still removed.
• Stabilize comment-only blocks by consistently placing comments attached to closing braces inside the body. This fixes the reported two-cycle Formatting did not converge failure without changing forced-break width accounting.
• Correct closing-brace indentation for tabular array initializers ending in line comments.
• Preserve line-comment text in flat layouts, including $NON-NLS$ and suppression markers, consistently with the existing no-reflow policy.
• Record the integrated-source fixes in the vendor patch.

Validation

Regression tests were added before the fixes and reproduced the reported failures. Coverage includes comment-only methods, constructors, initializers, type declarations, control-flow blocks and lambdas; array comments with and without trailing commas; and standalone JBang directives and trailing comment markers.

Import regression tests compile source before and after formatting and inspect the resulting field type for ordinary and static wildcard imports.

• ja test --all: 221 tests passed
• git diff --check: clean

Scope

No changes to structural wrapping policy are included. The style-evaluation document is excluded from this PR. Pre-existing drift between the vendor patch and the integrated visitor source is not reconciled here; this change adds only the correctness fixes to that patch.