fix(sqlmesh): repos should be partitioned by url not node_id

[ccerv1]

Since we are using urls as the source of truth, we should also fix this staging model. It is currently leading to a modest number of duplicate entries, which may have some downstream implications.

Here's an example:

SELECT * FROM stg_ossd__current_repositories
WHERE url = 'https://github.com/pulsarswap/landing'

Result:

[{
  "node_id":"R_kgDOOJZHZA",
  "id":949372772,
  "url":"https://github.com/pulsarswap/landing",
  "is_fork":false,
  "language":"TypeScript",
  "created_at":"2025-03-16 10:05:54.000 UTC", 
  "updated_at":"2025-03-17 16:33:15.000 UTC",
  "ingestion_time":"2025-10-26 18:08:50.494 UTC"
},{
  "node_id":"R_kgDOLA8IJA",
  "id":739182628,
  "url":"https://github.com/pulsarswap/landing",
  "is_fork":true,
  "language":"",
  "created_at":"2024-01-05 00:52:59.000 UTC",
  "updated_at":"2024-03-23 16:22:31.000 UTC",
  "ingestion_time":"2025-03-15 18:08:34.024 UTC"
}]

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
kariba-network	Ready	Preview	Comment	Oct 28, 2025 10:54am
oso-www	Ready	Preview	Comment	Oct 28, 2025 10:54am

[ravenac95]

I made #5540 as a follow on