fix: update foreign key constraints and ensure uniqueness in runner metadata

[johnfolly]

This pull request fixes several issues identified in COAL, specifically in the part responsible for sending data to PostgreSQL.

These issues are mainly related to the function that creates the link between the comsotech_runmetadata table and the result tables exported to PostgreSQL.

• First issue: there was a syntax error in the statement used to define the foreign key constraint. The ADD keyword was missing in the ALTER TABLE statement when declaring the constraint.

• Second issue: the foreign key constraint was being added to the wrong table name, without the expected prefix. For example, it targeted mytable instead of cosmotech_mytable by default.

• Third issue: the last_csm_run_id column was missing a uniqueness constraint, which caused the creation of the foreign key constraint to fail.

• Fourth issue: the code did not check whether the foreign key constraint already existed before attempting to create it. As a result, each write operation tried to recreate the constraint, which is not allowed by PostgreSQL.

• Fifth and final issue: when the same scenario was executed two or more times, data from the previous run was not deleted. This caused the insertion query into the runner_metadata table to fail because of the uniqueness constraint.

Each of these issues has been fixed and tested in a development environment.

However, some cases are still not handled, such as deleting a scenario, which currently does not remove the related data from the database.

One important remaining point is that we need to find an efficient way to test all these behaviors through integration tests directly in COAL.

[neomatamune]

Why do we put an overly complicated trigger when we should just do a drop cascade with the foreign keys on upsert (conflict should drop old row and insert new one)

[johnfolly]

Why do we put an overly complicated trigger when we should just do a drop cascade with the foreign keys on upsert (conflict should drop old row and insert new one)

I’ve simplified it

[lalepee]

Ok for me. I didn't found a better way to do the check on the existence of the foreign key constraint.