Added generalize fuzz testing for B-trees, fixed single-child bug

A single child is just another condition to watch out for during B-tree
merge, since a single-child obviously can't have a sibling.

This is a good safety to have, but I was surprised this can happen. But
it turns out to be quite easy since our rbyds defer the B-tree
operations until compaction. A merge down to a single child won't
propagate the merge until the parent compacts.

lfs.c

@@ -2232,6 +2232,7 @@ static int lfsr_rbyd_append(lfs_t *lfs, lfsr_rbyd_t *rbyd,
         other_id_ = id_;
     } else if (tag == LFSR_TAG_SHRINK) {
         LFS_ASSERT(id < rbyd->weight);
+        LFS_ASSERT(lfsr_data_len(data) <= rbyd->weight);
         // noop?
         if (lfsr_data_len(data) == 0) {
             return 0;
@@ -3320,8 +3321,6 @@ static lfs_ssize_t lfsr_btree_get(lfs_t *lfs,
             buffer, size);
 }
 
-// TODO drop the root during merges
-// TODO inline?
 static int lfsr_btree_commit(lfs_t *lfs,
         lfsr_btree_t *btree, lfs_size_t id,
         lfsr_rbyd_t *rbyd, const struct lfsr_attr *attrs) {
@@ -3821,6 +3820,11 @@ static int lfsr_btree_commit(lfs_t *lfs,
             goto abort;
         }
 
+        // only child? can't merge
+        if (pweight == parent.weight) {
+            goto abort;
+        }
+
         // last child? try the left sibling
         lfs_ssize_t sid;
         lfs_ssize_t sdelta;