fat zap should prefetch when iterating

When iterating over a ZAP object, we're almost always certain to iterate
over the entire object. If there are multiple leaf blocks, we can
realize a performance win by issuing reads for all the leaf blocks in
parallel when the iteration begins.

For example, if we have 10,000 snapshots, "zfs destroy -nv
pool/fs@1%9999" can take 30 minutes when the cache is cold. This change
provides a >3x performance improvement, by issuing the reads for all ~64
blocks of each ZAP object in parallel.

Reviewed-by: Andreas Dilger <andreas.dilger@whamcloud.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
External-issue: DLPX-58347
Closes #8862

include/sys/zap.h

@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
  * Copyright 2017 Nexenta Systems, Inc.
  */
 
@@ -350,6 +350,7 @@ typedef struct zap_cursor {
 	uint64_t zc_serialized;
 	uint64_t zc_hash;
 	uint32_t zc_cd;
+	boolean_t zc_prefetch;
 } zap_cursor_t;
 
 typedef struct {
@@ -375,7 +376,9 @@ typedef struct {
  * Initialize a zap cursor, pointing to the "first" attribute of the
  * zapobj.  You must _fini the cursor when you are done with it.
  */
-void zap_cursor_init(zap_cursor_t *zc, objset_t *ds, uint64_t zapobj);
+void zap_cursor_init(zap_cursor_t *zc, objset_t *os, uint64_t zapobj);
+void zap_cursor_init_noprefetch(zap_cursor_t *zc, objset_t *os,
+    uint64_t zapobj);
 void zap_cursor_fini(zap_cursor_t *zc);
 
 /*