Pipeline task management

Session persists a collection of tasks with unique identifiers; shared pipeline caller can optionally enqueue its call as a task instead of running it directly

model_server/base/pipelines/shared.py

@@ -47,7 +47,7 @@ class PipelineRecord(BaseModel):
 class PipelineQueueRecord(BaseModel):
     task_id: str
 
-
+# TODO: variant of this for queued tasks where accessor ID is separately parameterized
 def call_pipeline(func, p: PipelineParams) -> PipelineRecord:
     # match accessor IDs to loaded accessor objects
     accessors_in = {}