CMake and Meson build definitions are Out Of Sync

## Description

The CMake and Meson build definitions are maintained separately and have drifted. On current `main` (`e0399c6945326dd5281da07de8ed14a61c9f2cac`), equivalent configurations do not install the same public headers or run the same core tests.

Examples:

- CMake installs public headers by globbing each directory, while Meson lists them manually. As a result:
- CMake installs `iceberg/puffin_dv_io.h` and `iceberg/catalog/rest/auth/token_refresh_scheduler.h`, but Meson does not.
- Meson installs `iceberg/catalog/catalog_util.h`, `iceberg/catalog/session_catalog.h`, and `iceberg/catalog/session_context.h`, but CMake does not install headers from `src/iceberg/catalog/`.
- CMake includes `snapshot_summary_builder_test.cc` in `table_test`, but Meson does not include that test source, so the Meson CI path does not exercise it.
- Vendored/fallback dependency versions differ between the two build systems:

| Dependency | CMake | Meson |
|---|---:|---:|
| CRoaring | 4.4.3 | 4.3.11 |
| nlohmann/json | 3.11.3 | 3.12.0 |
| GoogleTest | 1.15.2 | 1.17.0 |

Other shared dependencies, such as nanoarrow 0.9.0, utf8proc 2.10.0, spdlog 1.15.3, cpr 1.14.1, and Google Benchmark 1.8.4, are currently aligned but can drift because their pins are also declared independently.
- Dependency resolution policy also differs: CMake requires system Zlib and Curl, while Meson can fall back to pinned wraps (Zlib 1.3.1 and Curl 8.10.1). Any intentional difference here should be documented and tested explicitly.

The Meson bundle gap is already tracked by #256 and is outside the scope of this issue. Other intentional feature differences should likewise remain explicit rather than appearing as accidental drift.

---

**This to track current status and initiated a discussion on how we can maintain and sync the build tools in the long term.**

Vergleiche die CMake- und Meson-Build-Definitionen im main-Commit e0399c6, beginnend mit der Installation öffentlicher Header, den table_test-Quellen und den aufgeführten Dependency-Pins. Prüfe snapshot_summary_builder_test.cc und Issue #256 auf den angegebenen Umfang. Als abgeschlossen gilt die Arbeit, wenn versehentliche Unterschiede angeglichen sind, während beabsichtigte Unterschiede bei Dependencies und Features dokumentiert und getestet sind.

Vom Indexierungsmodell aus dem Issue-Text verfasst.