Simplify and fix JobExportHistoryArchive / GenomeIndexToolData handling

Output parameters must behave somewhat like a HDA, in that they need to
have a dataset attribute that points to a Dataset instance.
`GenomeIndexToolData` fits that requirement, `JobExportHistoryArchive`
provides the fda property that behaves like a HDA.

lib/galaxy/jobs/__init__.py

@@ -1093,10 +1093,10 @@ class JobWrapper(HasResourceParameters):
         job = self._load_job()
 
         def get_special():
-            special = self.sa_session.query(model.JobExportHistoryArchive).filter_by(job=job).first()
-            if not special:
-                special = self.sa_session.query(model.GenomeIndexToolData).filter_by(job=job).first()
-            return special
+            jeha = self.sa_session.query(model.JobExportHistoryArchive).filter_by(job=job).first()
+            if jeha:
+                return jeha.fda
+            return self.sa_session.query(model.GenomeIndexToolData).filter_by(job=job).first()
 
         tool_evaluator = self._get_tool_evaluator(job)
         compute_environment = compute_environment or self.default_compute_environment(job)

lib/galaxy/tools/evaluation.py

@@ -86,7 +86,7 @@ class ToolEvaluator:
         if get_special:
             special = get_special()
             if special:
-                out_data["output_file"] = getattr(special, 'fda', None)
+                out_data["output_file"] = special
 
         # These can be passed on the command line if wanted as $__user_*__
         incoming.update(model.User.user_template_environment(job.history and job.history.user))