[pack #67] Don't pre-fetch packed objects during counting

This could be a real avenue, in single-threaded mode this probably
shouldn't happen and the work should rather be postponed to
entry generation, copying, which right now is ridiculously fast
as it doesn't have to do much work anymore.

> counting done 8.1M objects in 107.57s (75.5k objects/s)

Author: Byron

git-pack/src/data/output/count/objects.rs

@@ -463,7 +463,8 @@ fn id_to_count<Find: crate::Find>(
     statistics.expanded_objects += 1;
     output::Count {
         id: id.to_owned(),
-        entry_pack_location: db.location_by_oid(id, buf),
+        // entry_pack_location: db.location_by_oid(id, buf),
+        entry_pack_location: None,
     }
 }
 

gitoxide-core/src/pack/create.rs

@@ -210,6 +210,7 @@ where
         counts.shrink_to_fit();
         counts
     };
+    todo!("figure out performance issues");
 
     progress.inc();
     let num_objects = counts.len();