docs: clarify usage of upsert with surrogate primary keys

[laurenceisla]

Clarify the Upsert docs according to this conversation: #1118 (comment)

Also changed the example to one that uses natural primary keys instead.

[elyobo]

An example of how one would do this with surrogate keys as well would probably be helpful.

[wolfgangwalther]

Is the information in the issue still correct? I remember we added something about handling DEFAULT values instead of NULL for missing values at some point. Maybe that was not for POST in this case?

[laurenceisla]

Is the information in the issue still correct? I remember we added something about handling DEFAULT values instead of NULL for missing values at some point.

Thanks for the heads up, you're absolutely right! e.g. on GENERATED BY DEFAULT... or SERIAL PK columns we can add both preferences Prefer: missing=default, resolution=merge-duplicates and specify the columns to make this work (no need to link the issue anymore). I'll modify accordingly.

Also, I don't see an Upsert spec test that mixes both preferences, I'll add one too.

[laurenceisla]

@elyobo I added an example on what to do with surrogate keys. There is no need for a TRIGGER.

[wolfgangwalther]

I applied the docs commit on the v12 branch as well. The test commit didn't apply cleanly, didn't bother to fix that.

The docs commit could in theory be back-patched, further, because v11 seems to have missing=default as well - but again, it didn't apply cleanly, so I didn't bother.

At least we have the improved docs on the stable branch right now.