Plan typed table and incremental materialization operations

[dataders]

Summary

• resolve table and incremental control flow into typed, serializable operation plans
• distinguish logical Iceberg from Delta-backed UniForm and native managed Iceberg
• reject Delta-only partition overwrite for managed Iceberg and prevent managed Iceberg from selecting REPLACE ON
• retain explicit catalog provider, physical provider, DBR runtime/version, named capabilities, and selected SQL renderer variants
• plan config-change inclusion, insert-by-name, tags/catalog legality, column comments, constraints, and optimization before execution
• keep legacy Jinja fallbacks while ensuring typed leaf renderers do not probe provider, catalog, runtime, version, or capabilities

Dependency stack

• dbt-core: Execute semantic materialization strategies dbt-labs/dbt-core#16061
• dbt-adapters: Add semantic materialization lifecycle strategies dbt-labs/dbt-adapters#2139

Planning boundary

The plan owns decisions that change strategy, operation ordering, staging/replacement safety, overwrite behavior, operation inclusion, or SQL renderer variant. Relations, compiled SQL, resolved columns and predicates, and literal tag/property values remain late-bound execution inputs.

Validation

• full unit suite: 1,376 passed, 4 skipped
• final changed-surface suite: 191 passed, 2 skipped
• renderer tests deliberately fail if typed paths probe format, catalog, runtime capability, or re-plan creation
• Ruff fatal-error and unused-import checks passed; git diff checks passed
• live Databricks validation remains blocked by the invalid configured access token

This remains a draft while dependency PRs and live Databricks validation continue.