Fix: Dependency handling when converting on-run-start / on-run-end hooks in dbt projects

[izeigerman]

Previously, the conversion process didn't handle refs and sources in on-run-start / on-run-end hooks. This fix addresses the issue by sourcing ref and source dependencies directly from the manifest, in addition to extracting them by statically analyzing the hook's Jinja code.

[themisvaltinos]

the change looks good, but I noticed an issue in the _load_environment_statements which is my mistake that environment_statements is overwritten on each iteration of the loop so it doesn’t accumulate EnvironmentStatements across multiple projects. I believe to fix this we should replace the assignment with:

environment_statements.extend(
    statements
    for _, statements in sorted(
        hooks_by_package_name.items(),
        key=lambda item: 0 if item[0] == context.project_name else 1,
    )
)

and properly indent it. But if not addressed here, I can handle this in a subsequent PR

[izeigerman]

@themisvaltinos good catch. I've addressed your comment but keep in mind the following:

1. It appears that the dbt loader only supports a single dbt project today
2. Multiple dbt projects can have overlapping packages in which case we don't want the hooks to be duplicated. So I made hooks_by_package_name shared across projects instead.