Merge is not implicated.

I thought that the merge was implicated for making nodes too big, but now I'm convinced it's not.

When the brt tries to merge two nodes (maybe_merge_pinned_leaf_nodes) it has two cases
 1) The nodes combined are more than 3/4 a node, in which case it balances the nodes.
 2) Otherwise it merges them.  Hence merged nodes are always smaller than a node.

So I'm considering #1802 and #1811 to be fully resolved.

Refs #1802.  Closes #1811


git-svn-id: file:///svn/toku/tokudb@12836 c7de825b-a66e-492c-adef-691d508d4ae1

src/tests/inflate.c

@@ -7,7 +7,7 @@
  *    Close the file
  *    Insert some more to H (buffered in the root)
  *    Delete stuff from G (so that H merges with G)
- *     Merge first flushes G then merges the nodes, and maybe leaves G too big.
+ *    G ends up too big.
  */
 
 #include "test.h"

src/tests/inflate2.c

@@ -7,7 +7,7 @@
  *    Close the file
  *    Insert some more to H (buffered in the root)
  *    Delete stuff from G (so that H merges with G)
- *     Merge first flushes G then merges the nodes, and maybe leaves G too big.
+ *    G ends up too big (but it's not the merge)
  */
 
 #include "test.h"