Smart app launch and $everything operation should handle 403 error better

[d0roppe]

Describe the bug With the $everything operation, and having SMART app launch configured, a 403 error turns into a 500 error back to the caller.

Environment FHIR-main

To Reproduce Steps to reproduce the behavior: Call with /Patient//$everything with Oauth2 through SMART App launch a 403 error causes a 500 error back to the caller

Expected behavior the 403 error should flow back to the caller

[lmsurpre]

with the changes from #3779, the expected response status for this case is now a 403 (not a 500) and the error message should be like this:

Error retrieving $everything resources of type 'Appointment' for Patient/002e2eac-c802-41fc-8b38-6d45b67a38d0; caused by read permission for 'Appointment' is not granted by any of the provided scopes: [[patient/Patient.*]]

[lmsurpre]

During QA, we found that we still emitted a generic 500 error in the case that the user was missing read access to their patient resource (e.g. missing scope patient/Patient.read).

[lmsurpre]

After the fix, we confirmed some additional cases to ensure we're not leaking information, and everything looks OK to us.