Known issues with dbt-fusion

dbt Fusion is very particular about things, in a way that causes problems for how the harness is set up. These are the known issues:

### Failures can fail hard

Fusion will not run if there are references to models that don't exist. Since the point of ADE-bench is to see if an agent can create these models—and in some cases they won't, and they won't be there—tests that check for the existence of these models will prevent dbt from compiling. This means that correct solutions will pass, but incorrect solutions fail harder than they should (i.e., instead of dbt tests failing, the entire project will not parse).

### Seed issues with `\`.

For reasons I don't yet understand, running `dbt seed` against Snowflake with fusion removes `\` from text fields. Prior versions of dbt don't do this. So several Airbnb tasks fail, because of this issue.

Affected tasks:
- None, currently

### There is no `raw_sql` node

Fusion does not return the [raw sql attribute](https://docs.getdbt.com/docs/fusion/supported-features#:~:text=You%20access%20the,in%20the%20future.). Several tests depend on this, so these tests will not work.

Affected tasks:
- airbnb006
- quickbooks002
- quickbooks003
- quickbooks004