[Feature] Roadmap for Paimon C++ 0.4.0

### Search before asking

- [x] I searched in the [issues](https://github.com/apache/paimon-cpp/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

- [x] Support shared-shredding columnar storage for `MAP` based on [PIP-43](https://cwiki.apache.org/confluence/spaces/PAIMON/pages/430408347/PIP-43+Columnar+Storage+Optimization+for+MAP+Type+in+Paimon), including adaptive physical columns, field mappings, overflow storage, and end-to-end reads and writes.
- [ ] Support richer data types, starting with `VECTOR` based on [PIP-40](https://cwiki.apache.org/confluence/spaces/PAIMON/pages/399279132/PIP-40+Introduce+a+new+Vector+data+type), 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](https://cwiki.apache.org/confluence/spaces/PAIMON/pages/444334302/PIP-46+Support+pluggable+real-time+writes+and+memory+disk+union+reads+for+Paimon+C), tracked by [#158](https://github.com/apache/paimon-cpp/issues/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](https://github.com/apache/paimon-cpp/issues/174).
- [ ] Support Format Table reads and writes for Hive-style file directories, tracked by [#170](https://github.com/apache/paimon-cpp/issues/170).
- [ ] Extend Data Evolution to work with deletion vectors, tracked by [#169](https://github.com/apache/paimon-cpp/issues/169).
- [ ] Extend Data Evolution to primary-key tables and support compaction across evolved field groups, tracked by [#204](https://github.com/apache/paimon-cpp/issues/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?

- [x] I'm willing to submit a PR!

Commencez par les issues ciblées liées, en particulier #158 pour les écritures en temps réel et les lectures d’union, #174 pour la production du changelog, #170 pour Format Table, et #169 ou #204 pour Data Evolution. Lisez les documents PIP référencés lorsque cela s’applique et choisissez un élément de roadmap dont le périmètre est défini. La définition de terminé exige une modification ciblée et révisable, avec les tests unitaires et d’intégration pertinents, la documentation et la validation de la compatibilité.

Rédigé par le modèle d'indexation à partir du texte de l'issue.