Improve retry logic to handle invalid cache and no changes to Observation

[namalu]

This PR includes updates to the R4FhirImportService SaveObservationAsync logic.

Invalid cache entry:

If an Observation has been saved to the in memory cache and deleted from the FHIR server, it may take an hour for the cache to be updated. During this time, when new data is sent, PUT operations on the deleted Observation will fail. These code updates will handle the failure by removing the invalid Observation from the cache, creating a new Observation in the FHIR server, and saving the new Observation to the cache.

Observations updated repeatedly:

If a batch of data contains one or more messages that can successfully be transformed into Observations, but also contains one or more messages that will cause an "unhandled" exception, the entire batch will be retried indefinitely. The retry will include updating the successful Observations in the FHIR Server repeatedly. These code updates will check the new Observation with the existing Observation and only perform a PUT if there is a change. This will prevent unnecessary updates to Observations, and reduce requests to the FHIR Server.

[dustinburson]

Added some suggestions to optimize but overall the changes look good.