Add sampling fields to dataset descriptor (facebookresearch#3782)

Summary:
Pull Request resolved: facebookresearch#3782

Fields sampling_column and sampling_rate are added to dataset descriptor for supporting sampling in dataset readers.

Reviewed By: satymish

Differential Revision: D61569067

fbshipit-source-id: e5db9957538b033bbef4b7662154411b9044d1f8

Author: kuarora

benchs/bench_fw/descriptors.py

@@ -80,6 +80,11 @@ class DatasetDescriptor:
 
     embedding_column: Optional[str] = None
 
+    sampling_rate: Optional[float] = None
+
+    # sampling column for xdb
+    sampling_column: Optional[str] = None
+
     def __hash__(self):
         return hash(self.get_filename())