previous implementation used C.memmove to do heavy lifting
doing this involves passing a pointer into the []byte to C
in Go 1.6 this has an unintended performance penalty as it
unnecessarily checks the entire slice

this is a bug in Go 1.6 documented here:
golang/go#14387

however, it seems as though this call can be avoided entirely,
by creating an artificial slice header which points to the
newly malloc'd C buffer, we use Go's builtin copy() method
in addition to avoid this bug, it also avoids a cgo call
which we know can add up

Change-Id: I910e5705e29e6c86343fd0e6315c8430f56b07ef
Reviewed-on: http://review.couchbase.org/60991
Reviewed-by: Steve Yen <steve.yen@gmail.com>
Tested-by: Marty Schoch <marty.schoch@gmail.com>

It was discovered in #64 that void* arguments cause allocations in cgo
and extra time checking arguments as opposed to char* arguments.

The issue golang/go#14387 claims that it will fix this problem but I am
going to keep this patch around for testing against tip and later
merging if necessary.

It was discovered in #63 that void* arguments cause allocations in cgo
and extra time checking arguments as opposed to char* arguments.

The issue golang/go#14387 claims that it will fix this problem but I am
going to keep this patch around for testing against tip and later
merging if necessary.

Fixes #14387.

Change-Id: Icc98be80f549c5e1f55c5e693bfea97b456a6c41
Reviewed-on: https://go-review.googlesource.com/19621
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-on: https://go-review.googlesource.com/22036
Reviewed-by: Ian Lance Taylor <iant@golang.org>