feat: introduce data type with JSON serialization

[lszskye]

Purpose

Introduce the Paimon data type system for apache-paimon-cpp, including:

• DataType: Base class representing Paimon data types, with Arrow type mapping and JSON serialization. Supports atomic types (BOOLEAN, TINYINT, SMALLINT, INT, BIGINT, FLOAT, DOUBLE, STRING, BYTES, DATE, DECIMAL, TIMESTAMP, BLOB) and complex types (ARRAY, MAP, ROW).
• DataField: Represents a field in a row type with id, name, type, and optional description. Provides bidirectional conversion between Paimon DataField and Arrow Field/Schema, including metadata handling for field ids.
• RowType: A structured type containing named fields, serialized as JSON with field definitions.
• ArrayType: A list/array type with element type.
• MapType: A key-value map type. Note: keys are always non-nullable due to Apache Arrow limitations.
• RowKind: Represents row kinds (INSERT, UPDATE_BEFORE, UPDATE_AFTER, DELETE) with byte value and short string serialization.
• DataTypeJsonParser: A comprehensive type parser supporting the full Paimon type syntax (including parameterized types like DECIMAL(10,2), TIMESTAMP(9) WITH LOCAL TIME ZONE, etc.).

Tests

• DataTypeTest
• DataFieldTest
• DataTypeJsonParserTest
• RowKindTest

[leaves12138]

Thanks for the contribution. I found one parser correctness issue that should be fixed before merge.

Tokenize is missing a break after handling CHAR_END_SUBTYPE (>). This makes every > token also emit a BEGIN_PARAMETER (() token at the same position. As a result, SQL-style nested type strings using subtype syntax, such as ROW<id INT> or ARRAY<INT>, cannot be parsed correctly because the token stream contains an extra ( after each >.

Please add the missing break and include a regression test for a type string containing >.

[leaves12138]

Re-reviewed the latest update. The previously reported CHAR_END_SUBTYPE fallthrough has been fixed. I did not find further blockers in this round.