🤖 backported "Keep collection_id of dashboard subscriptions in sync with same field on dashboard"

[github-actions]

#17537

[noahmoss]

@flamber this is now merged to master -- if you have time to try it out soon and determine whether it's safe to include in the 40.3 release, that would be great :)

[codecov]

Codecov Report

Merging #17586 (55c4ca9) into release-x.40.x (ca559f4) will increase coverage by 0.06%.
The diff coverage is 88.88%.

@@                Coverage Diff                 @@
##           release-x.40.x   #17586      +/-   ##
==================================================
+ Coverage           85.01%   85.07%   +0.06%     
==================================================
  Files                 416      416              
  Lines               33291    33317      +26     
  Branches             2377     2379       +2     
==================================================
+ Hits                28301    28346      +45     
+ Misses               2613     2592      -21     
- Partials             2377     2379       +2     

Impacted Files	Coverage Δ
src/metabase/cmd.clj	42.57% <50.00%> (+13.86%)	⬆️
src/metabase/models/pulse.clj	87.41% <73.68%> (-0.70%)	⬇️
...end/src/metabase_enterprise/serialization/load.clj	85.07% <87.50%> (+0.07%)	⬆️
...kend/src/metabase_enterprise/serialization/cmd.clj	84.61% <90.00%> (+6.39%)	⬆️
...d/src/metabase_enterprise/serialization/upsert.clj	86.74% <100.00%> (+2.65%)	⬆️
src/metabase/api/field.clj	84.18% <100.00%> (+0.54%)	⬆️
src/metabase/models/card.clj	94.59% <100.00%> (ø)
src/metabase/models/dashboard.clj	82.12% <100.00%> (+0.15%)	⬆️
src/metabase/models/metric.clj	90.76% <100.00%> (+0.44%)	⬆️
src/metabase/models/permissions_group.clj	81.81% <100.00%> (-0.73%)	⬇️
... and 14 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7f052ce...55c4ca9. Read the comment docs.

[flamber]

This works as expected for all three appdb types. I would probably just have used the H2 SQL for all, since it's the lowest denominator and works for others too.

[noahmoss]

@flamber Thanks for reviewing!

I originally had the nested query for all DBs, but Cam was concerned about its efficiency if there are a large number of pulses.