sparse-read: discard gap below 65K only

Testing on actual data shows that 65K is more efficient in both time and memory
than 256K.

mercurial/configitems.py

@@ -597,7 +597,7 b" coreconfigitem('experimental', 'sparse-r"
     default=0.50,
 )
 coreconfigitem('experimental', 'sparse-read.min-gap-size',
-    default='256K',
+    default='65K',
 )
 coreconfigitem('experimental', 'treemanifest',
     default=False,