fix: separate post-commit invalidation callbacks from trigger and user callbacks

[wschurman]

Why

The issue we're seeing is:

• A post-commit trigger is calling load on data that was deleted. This call should be possible (using loadNullable).
• The cache still has data in it since the ordering of post-commit callbacks is undefined.
• The data is loaded even though it no longer exists.

How

To fix this, add some strict ordering to post-commit callbacks to ensure that post-commit invalidations run before any other callbacks. This ensures that the post-commit callbacks that access the cache have consistent cache loads.

Test Plan

• Run the new test without the fix, see that it fails (the exception is thrown)
• Run the new test with the fix, see that it passes.

[codecov]

Codecov Report

Merging #144 (ba45e80) into master (c4df889) will increase coverage by 0.01%.
The diff coverage is 100.00%.

❗ Current head ba45e80 differs from pull request most recent head fcf5dc2. Consider uploading reports for the commit fcf5dc2 to get more accurate results

@@            Coverage Diff             @@
##           master     #144      +/-   ##
==========================================
+ Coverage   95.89%   95.90%   +0.01%     
==========================================
  Files          73       73              
  Lines        1851     1857       +6     
  Branches      228      205      -23     
==========================================
+ Hits         1775     1781       +6     
  Misses         75       75              
  Partials        1        1              

Flag	Coverage Δ
integration	95.90% <100.00%> (+0.01%)	⬆️
unittest	95.90% <100.00%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
packages/entity/src/EntityMutator.ts	98.20% <100.00%> (ø)
packages/entity/src/EntityQueryContext.ts	100.00% <100.00%> (ø)

---

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 c4df889...fcf5dc2. Read the comment docs.

[ide]

Sounds right. Triggers logically don't care about the cache. There still is a race condition where another process reading the same data at this time could fill the cache, but we have that problem anyway.