crypto/sha512: block implementation in amd64 assembly

Benchmark on Intel(R) Xeon(R) CPU X5650  @ 2.67GHz

benchmark              old ns/op    new ns/op    delta
BenchmarkHash8Bytes         1779         1114  -37.38%
BenchmarkHash1K             9848         4894  -50.30%
BenchmarkHash8K            68513        32187  -53.02%

benchmark               old MB/s     new MB/s  speedup
BenchmarkHash8Bytes         4.50         7.18    1.60x
BenchmarkHash1K           103.97       209.19    2.01x
BenchmarkHash8K           119.57       254.51    2.13x

R=agl
CC=golang-codereviews
https://golang.org/cl/37150044

src/pkg/crypto/sha512/sha512block.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+// +build !amd64
+
 // SHA512 block step.
 // In its own file so that a faster assembly or C version
 // can be substituted easily.

src/pkg/crypto/sha512/sha512block_decl.go

+// Copyright 2013 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build 386 amd64
+
+package sha512
+
+//go:noescape
+
+func block(dig *digest, p []byte)