fix(postgres): make advisory lock cancel safe

[joeydewaal]

This pr makes the PgAdvisoryLock::acquire cancel safe by wrapping the connection in a PgAdvisoryLock before actually acquiring the lock. If the acquire future is dropped the drop impl will release the lock.

Does your PR solve an issue?

fixes #4198

Is this a breaking change?

Nope

[joeydewaal]

Note: I've decided to not update the try_acquire implementation because releasing the lock depends on the result of the query. So that method is still not cancel safe unfortunately.

[abonander]

Note: I've decided to not update the try_acquire implementation because releasing the lock depends on the result of the query. So that method is still not cancel safe unfortunately.

Can you just update the documentation to explain that? And add a Cancel Safety clause to the regular lock method that explains what it'll do on cancel?

[abonander]

Thanks!