Fix deployment idempotency for dimension links and downstream validation

[shangyian]

Summary

Dimension Link Reshuffling

When a node has multiple links to the same dimension with no role (e.g., two date links with different join_on), the matching logic used (dimension name, role) as the identity key. This caused it to treat distinct links as duplicates (deleting one and updating the other) on every push, which triggered downstream revalidation.

After session.flush() in bulk_deploy_nodes_in_level, SQLAlchemy expired the dimension links collection on new revisions. Later access in _deploy_links triggered a lazy load that silently failed in async context, causing the match to find 0 existing links and create duplicates.

Impact Propagation

Freshly deployed nodes had raw string column types (e.g., 'bigint') that hadn't round-tripped through ColumnTypeDecorator. When impact propagation passed these to validate_node_query, it would fail because it expected parsed ColumnType objects.

Test Plan

• PR has an associated issue: #
• make check passes
• make test shows 100% unit test coverage

Deployment Plan

[netlify]

✅ Deploy Preview for thriving-cassata-78ae72 canceled.

Name	Link
🔨 Latest commit	422fff1
🔍 Latest deploy log	https://app.netlify.com/projects/thriving-cassata-78ae72/deploys/69e616293dff7900078bc979