Bug: Attempting to do a versioned read of a resource that was just expunged causes NullPointerException until it is expired in memory cache

[ttntrifork]

Describe the bug

• Attempting to do a versioned read of a resource that was just expunged (via MdmStorageInterceptor when isAutoExpungeGoldenResources == true) causes NullPointerException, until it is expired in the "forced id" memory cache (1 minute expiration), after which it then correctly throws ResourceNotFoundException as expected.

Expected behavior

• The versioned read should cause a ResourceNotFoundException even if the PID is not yet expired in the memory cache

To Reproduce

• I will open a PR that simulates the problem via a unit test and fixes it by introducing a null guard such that BaseHapiFhirResourceDao correctly throws a ResourceNotFoundException even if the PID is not yet expired in the memory cache

---

Fix Details
The problem could probably also be fixed by invalidating the caches after commit, as is done in https://github.com/hapifhir/hapi-fhir/blob/v8.4.0/hapi-fhir-jpaserver-base/src/main/java/ca/uhn/fhir/jpa/dao/expunge/JpaResourceExpungeService.java#L212-L225
But the null guard seems good to have in any case (and fixes the issue without invalidating the entire cache).

• It could even be considered that the JpaResourceExpungeService should stop invalidating the cache (and thereby stop incurring a performance cost) with this fix?

Stacktrace:

Caused by: java.lang.NullPointerException: Cannot invoke "ca.uhn.fhir.jpa.model.entity.BaseHasResource.getVersion()" because "entity" is null
	at ca.uhn.fhir.jpa.dao.BaseHapiFhirResourceDao.readEntity(BaseHapiFhirResourceDao.java:1944)
	at ca.uhn.fhir.jpa.dao.BaseHapiFhirResourceDao.lambda$readEntity$21(BaseHapiFhirResourceDao.java:1680)
	at ca.uhn.fhir.jpa.dao.tx.HapiTransactionService.invokeCallableAndHandleAnyException(HapiTransactionService.java:663)
	... 13 more