ARROW-11745: [C++] Add helper to generate random record batches by schema

[lidavidm]

This adds a vaguely Hypothesis-esque helper to generate a random record batch from a list of fields, whose metadata is inspected and used to set generation parameters (e.g. min/max value).

[github-actions]

https://issues.apache.org/jira/browse/ARROW-11745

[bkietz]

This is a huge improvement, thanks!

I think it'd be worthwhile to add an overload of arrow::field which takes metadata but not nullable:

diff --git a/cpp/src/arrow/type_fwd.h b/cpp/src/arrow/type_fwd.h
index 230c1ff6c..578aa2d63 100644
--- a/cpp/src/arrow/type_fwd.h
+++ b/cpp/src/arrow/type_fwd.h
@@ -629,6 +629,18 @@ std::shared_ptr<Field> ARROW_EXPORT
 field(std::string name, std::shared_ptr<DataType> type, bool nullable = true,
       std::shared_ptr<const KeyValueMetadata> metadata = NULLPTR);
 
+/// \brief Create a Field instance
+///
+/// The field will be nullable.
+///
+/// \param name the field name
+/// \param type the field value type
+/// \param metadata any custom key-value metadata
+inline std::shared_ptr<Field> field(std::string name, std::shared_ptr<DataType> type,
+                                    std::shared_ptr<const KeyValueMetadata> metadata) {
+  return field(std::move(name), std::move(type), /*nullable=*/true, std::move(metadata));
+}
+
 /// \brief Create a Schema instance
 ///
 /// \param fields the schema's fields

[bkietz]

I think this is basically ready to go, just a few more comments:

[bkietz]

👍

[bkietz]

LGTM, merging

[emkornfield]

Any plans of applying this to parquet roundtrip tests?

[lidavidm]

@emkornfield could be done, did you want to file a JIRA for that? (Looks like the current tests all use hardcoded data?)