enable share_common_data for "isabelle dump" and its derivatives (e.g. "isabelle mmt_import"): this has the potential to reduce ML heap size considerably, after initial command definitions;

--- a/src/Pure/Tools/dump.scala	Wed Aug 28 22:59:49 2019 +0200
+++ b/src/Pure/Tools/dump.scala	Wed Aug 28 23:01:53 2019 +0200
@@ -171,6 +171,7 @@
       val use_theories_result =
         session.use_theories(use_theories,
           unicode_symbols = unicode_symbols,
+          share_common_data = true,
           progress = progress,
           commit = Some(Consumer.apply _))