feat(athena): add model-level timeout support

[dtaniwaki]

Summary

• Add model_timeout_seconds (model config) to limit total model execution time including batch loops
• Thread-safe implementation using threading.local() for deadline tracking
• Check timeout before each batch iteration and in _run_query()
• Raise AthenaModelTimeoutError when the deadline is exceeded

Motivation

The dbt-athena adapter has no model-level timeout mechanism. When Iceberg force_batch generates a large number of batches, each batch query is fast but the total model execution time becomes unbounded. This makes it impossible to set a reasonable CodeBuild timeout without risking runaway models.

Standardizing timeout handling is being discussed upstream at dbt-labs/dbt-core#8522, but progress has stalled with no concrete implementation plan. This PR provides a practical solution for the Athena adapter in the meantime.

Usage

In dbt_project.yml:

models:
  my_project:
    my_heavy_model:
      +model_timeout_seconds: 3600  # 1 hour

Or in a model file:

{{ config(model_timeout_seconds=3600) }}
SELECT ...

Changes

• exceptions.py: Add AthenaModelTimeoutError
• impl.py: Add model_timeout_seconds to AthenaConfig, add set/check/clear_model_timeout() adapter methods with threading.local(), add timeout check in _run_query() and run_operation_with_potential_multiple_runs()
• table.sql, incremental.sql: Set/clear model timeout at materialization boundaries
• create_table_as.sql, helpers.sql, merge.sql: Check model timeout before each batch iteration

Test plan

• Unit tests for model timeout lifecycle (set, check, clear)
• Thread isolation test (independent deadlines per thread)
• Edge cases (None, 0, negative values are no-ops)
• Existing unit tests pass

Recreated from #1, rebased on main for cleaner history.