TableMetadata is not thread-safe

### Apache Iceberg version

1.11.0 (latest release)

### Query engine

Trino

### Please describe the bug 🐞

`RESTSessionCatalog`'s `RESTTableCache` caches `TableMetadata` and shares it across threads. `TableMetadata` itself handles this correctly - its lazy fields use `synchronized` + `volatile`. However, `BaseSnapshot` and `Schema`, reachable from `TableMetadata`, have lazily-initialized fields with no synchronization:

https://github.com/apache/iceberg/blob/7aa4f97fae8f6393950d925ca67e8331a780a20a/core/src/main/java/org/apache/iceberg/BaseSnapshot.java#L51-L57

https://github.com/apache/iceberg/blob/7aa4f97fae8f6393950d925ca67e8331a780a20a/api/src/main/java/org/apache/iceberg/Schema.java#L77-L83

`PartitionSpec` follows the same lazy pattern but correctly uses `synchronized` + `volatile`.

### Willingness to contribute

- [ ] I can contribute a fix for this bug independently
- [ ] I would be willing to contribute a fix for this bug with guidance from the Iceberg community
- [ ] I cannot contribute a fix for this bug at this time

Start with the lazy fields at the referenced lines in core/src/main/java/org/apache/iceberg/BaseSnapshot.java and api/src/main/java/org/apache/iceberg/Schema.java, then compare their initialization pattern with PartitionSpec. Trace how RESTTableCache shares TableMetadata across threads and verify that concurrent access to these fields is safe without changing the intended behavior.

Written by the indexing model from the issue text.