fix: When looking for stray triggers ignore partitions

[ponychicken]

Addresses #56

Partitions of tables are not considered when looking for additional triggers or cleaning them up (is handled by PG through the root table)

[cheerfulstoic]

Thanks very much!

I modified the query to use joins because that seemed cleaner to me. More importantly, perhaps, is that I changed the tests to CREATE OR REPLACE TRIGGER on partitioned_root instead of partitioned_part1 because you would be (if I'm thinking about this right) putting the ecto_watch watcher on the partition root and therefore adding / removing triggers via that root table, which as you shared causes the partitioned tables to add triggers. The documentation didn't seem to say if a trigger is dropped from the root that the trigger is dropped from the partitions, but I just tested and it seems that is the case.

Please have a look at the most recent commit and, if it looks good, I'll release a new version:

b0c892a

[ponychicken]

Yes, you are right it makes more sense to install the test trigger on the root table and check that there is no warning about the partition trigger. I ran your new query on the DB here and it correctly returns the root trigger but not the partition triggers.

[ponychicken]

And yes, I observe the same behaviour, when I delete the root trigger the partition triggers are gone as well. (PG17)

[cheerfulstoic]

Released as 0.15.0, thanks for your help!