refactor(reader): make VarBinArray non-sealed, split its modes out

[dfa1]

Summary

• Follows up on a design question raised while reviewing fix(reader): make constant Utf8/Binary decode lazy #331: VarBinArray was the one Array sub-hierarchy still declared sealed with an explicit permits list, unlike its siblings (ByteArray, LongArray, DecimalArray, ...) which are non-sealed and let each representation live as an ordinary top-level class.
• Grepped the codebase for exhaustive switch/pattern-matching over VarBinArray's specific modes — zero matches (only one instanceof VarBinArray.OffsetMode check). So the sealing bought no compiler-enforced exhaustiveness, only coupling the interface's declaration to the full list of its implementations, and forcing every new representation to be nested inside VarBinArray.java and added to its permits clause (as ConstantMode was in fix(reader): make constant Utf8/Binary decode lazy #331).
• VarBinArray is now non-sealed. Each former nested record is promoted to a top-level class in reader.array, renamed to match the sibling families' self-describing naming:
  • OffsetMode → VarBinOffsetArray
  • DictMode → VarBinDictArray
  • ChunkedMode → VarBinChunkedArray
  • ViewMode → VarBinViewArray
  • SlicedMode → VarBinSlicedArray
  • ConstantMode → VarBinConstantArray
• checkedLength (bounds-check shared by the offset- and dict-backed representations) moved to a new package-private VarBinArrays helper, mirroring the RunEndArrays precedent.
• Pure rename/move, no behavior change. docs/compatibility.md and CHANGELOG.md updated.

Breaking for any code referencing the old nested type names (VarBinArray.OffsetMode etc.) directly.

Test plan

• ./mvnw verify -DskipITs (whole repo)
• ./mvnw -pl integration -am verify -Dit.test=DocsConsistencyTest -Dvortex.it.excludedGroups=

🤖 Generated with Claude Code