[Feature] Roadmap for Paimon C++ 0.4.0

从链接的聚焦 issue 开始，尤其是关于实时写入和 union 读取的 #158、关于 changelog 生成的 #174、关于 Format Table 的 #170，以及关于 Data Evolution 的 #169 或 #204。在适用的情况下阅读所引用的 PIP 文档，并选择一个范围明确的 roadmap 项目。完成的要求是进行一项聚焦且可供 review 的更改，并包含相关的 unit 和 integration 测试、文档以及兼容性验证。

Search before asking

• I searched in the issues and found nothing similar.

Motivation

Looking ahead, Paimon C++ will focus on several areas:

• Storage and retrieval optimizations for vertical workloads, including efficient MAP storage, richer data types such as VECTOR, and complete index support.
• Making newly written data queryable in real time through pluggable real-time writes and memory/disk union reads.
• Following up on important capabilities from the broader Paimon community, including changelog production, Format Table, and extensions to Data Evolution.

Solution

1. Storage and retrieval optimizations for vertical workloads

• Support shared-shredding columnar storage for MAP<STRING, T> based on PIP-43, including adaptive physical columns, field mappings, overflow storage, and end-to-end reads and writes.
• Support richer data types, starting with VECTOR<T, N> based on PIP-40, including schema representation, storage, reads, writes, and Data Evolution.
• Complete File Index support by adding index generation to the existing read path.
• Continue improving storage layout, predicate pushdown, column pruning, point lookups, and index-assisted retrieval for workload-specific scenarios.

2. Making real-time data queryable

• Support pluggable real-time writes and memory/disk union reads based on PIP-46, tracked by #158.
• Make data held by an active writer queryable before it is committed into a snapshot.
• Guarantee consistent query results without missing or duplicate rows while writing, committing, and reclaiming memory segments concurrently.
• Support append tables first, followed by primary-key tables, deletion-vector mode, and additional production scenarios.

3. Following up on Paimon community capabilities

• Support changelog production, including input, lookup, and full-compaction modes, tracked by #174.
• Support Format Table reads and writes for Hive-style file directories, tracked by #170.
• Extend Data Evolution to work with deletion vectors, tracked by #169.
• Extend Data Evolution to primary-key tables and support compaction across evolved field groups, tracked by #204.
• Continue following relevant Paimon features and make them available through native Paimon C++ APIs where appropriate.

Each roadmap item should be delivered through focused issues and reviewable pull requests, with unit tests, integration tests, documentation, and compatibility validation where applicable.

Anything else?

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!