[XiaoHongbo-Hope]

Add database-scoped Model metadata APIs to Java and PyPaimon.

What changed

• Add create, get, list, and drop Model metadata APIs.
• Define the shared REST/OpenAPI contract and DLF action mapping.
• Store provider, provider model name, input schema, optional fixed outputs, and provider properties.
• Validate Model fields consistently in Java and Python.
• Support idempotent create and drop through ignore_if_exists and ignore_if_not_exists.
• Omit absent optional fields and serialize empty model lists as [].
• Keep prompts, credentials, and prediction execution outside Model metadata.

Design references

• Apache Flink Model DDL: Model is scoped by Catalog and Database.
• BigQuery Model resource: Model is a Dataset-scoped metadata resource.
• Numeric versioning is deferred until update and lifecycle semantics are defined.

Prediction submission is deferred to a follow-up.

Tests

• Java paimon-api: 270 passed
• Python Model tests: 5 passed
• Non-fast paimon-core reactor compile: BUILD SUCCESS