feat: Add Arrow JSON integration testing executable. Fix ArrowReader bug.

[willtemperley]

What's Changed

Added an Arrow-JSON integration testing executable as discussed in #96. This includes an ArrowJSON struct for decoding and encoding the Arrow JSON test files.

Added a rule to .swiftlint.yml to ignore case statements in cyclomatic complexity checks.

Made some Arrow Type declarations public to support encoding and decoding.

Fix: ArrowReader now creates a null buffer of the length specified by the flatbuffers definition, instead of calculating it from the array length. This fixes a bug where a null buffer of full length would be created and allocated as zeros, making a non-null array become all-null.

[maltzsama]

I tested the integration test executable on Debian 13 against generated_primitive from apache/arrow-testing (cpp-21.0.0 gold files).
What works:

• ARROW_TO_JSON
• VALIDATE (on the .arrow_file format)
• Build compiles cleanly (only deprecation warnings from flatbuffers and ArrowCExporter)
  What doesn't work yet (not a blocker for this PR):
• JSON_TO_ARROW and VALIDATE on .arrow_file fail with:

$ swift run arrow-json-integration-test \                                                                                                                            base 15:40:04
  --arrow="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.arrow_file" \
  --json="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.json" \
  --mode=ARROW_TO_JSON
Building for debugging...
[1/1] Write swift-version-7FB2667E20D8AC03.txt
Build of product 'arrow-json-integration-test' complete! (0.28s)
$ swift run arrow-json-integration-test \                                                                                                                            base 15:40:09
  --arrow="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.arrow_file" \
  --json="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.json" \
  --mode=JSON_TO_ARROW
Building for debugging...
[1/1] Write swift-version-7FB2667E20D8AC03.txt
Build of product 'arrow-json-integration-test' complete! (0.28s)
Error: DecodingError.typeMismatch: expected value of type DataValue. Path: batches[0].columns[0].DATA[0]. Debug description: Cannot decode DataValue
$ swift run arrow-json-integration-test \                                                                                                                            base 15:40:16
  --arrow="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.arrow_file" \
  --json="/tmp/arrow-testing/data/arrow-ipc-stream/integration/cpp-21.0.0/generated_primitive.json" \
  --mode=VALIDATE     
Building for debugging...
[1/1] Write swift-version-7FB2667E20D8AC03.txt
Build of product 'arrow-json-integration-test' complete! (0.29s)
Error: DecodingError.typeMismatch: expected value of type DataValue. Path: batches[0].columns[0].DATA[0]. Debug description: Cannot decode DataValue

But I believe this DataValue in ArrowJSON.swift decode can be addressed in another issue.

@kou LGTM

[kou]

Should we use toFile() instead of writeFile() to write the ARROW1 marker?

[kou]

How about comparing in Apache Arrow data instead of JSON data?
Archery doesn't use the Apache Arrow -> JSON conversion. If we compare data in Apache Arrow, we don't need to implement Apache Arrow -> JSON conversion. The Apache Arrow data comparison will be also useful in other use cases.