feat(sfdc_formula_view): add query_engine inference and use run_query for non-Redshift adapters

[fivetran-jacklowery]

Summary

This PR updates sfdc_formula_view to select the generated formula model by query engine and avoid a relation-existence check that can fail for Snowflake catalog-linked databases.

The Salesforce connector produces multiple rows in fivetran_formula_model for the same Salesforce object — one per query engine. The previous macro filtered only by object and used dbt_utils.get_column_values, whose load_relation check can return None for catalog-linked (Iceberg/Managed Data Lake) sources even when direct SQL against the source succeeds.

Changes:

• Adds an optional query_engine argument to sfdc_formula_view
• Adds a salesforce_formula_query_engine project var as a fallback
• Infers query engine from target.type if neither is provided (snowflake → Snowflake, bigquery → BigQuery, databricks/spark → Databricks, redshift → Redshift, else Generic)
• Replaces dbt_utils.get_column_values on the non-Redshift path with a direct run_query call that filters by both object and query_engine
• Preserves all existing Redshift model_large / SUPER behavior unchanged
• Preserves the materialization argument; documents that catalog-linked Snowflake databases require materialization='table' because view creation is unsupported

Validation

Smoke-tested against a Snowflake catalog-linked database backed by Fivetran Managed Data Lake using fivetran/dbt_snowflake_iceberg_writer:

dbt run --select smoke_salesforce_formula_account smoke_salesforce_formula_user_role
# 2 pass, 0 fail

Both models materialized as BASE TABLE relations against lowery_linked_writer.dbt_core_dev.

Test plan

• Confirm dbt parse passes with updated macro
• Confirm dbt compile --select smoke_salesforce_formula_account smoke_salesforce_formula_user_role passes
• Confirm dbt run --select smoke_salesforce_formula_account smoke_salesforce_formula_user_role passes with materialization='table' on Snowflake catalog-linked database
• Confirm existing Redshift integration tests still pass
• Add/update integration tests: multiple query_engine rows for same object selects adapter-specific row; explicit query_engine override works; default adapter inference works

🤖 Generated with Claude Code