Also add a call to ioutil.ReadAll to mimic the use in revid. There is
not a great deal of difference on the RPi3 in terms of throughput,
though the allocations obviously increase.
```
$ go test -run ^$ -bench . -benchmem -benchtime 30s
goos: linux
goarch: arm
BenchmarkRoundTrip-4 2000 20184818 ns/op 1.49 MB/s 104646 B/op 4 allocs/op
BenchmarkRoundTripWriterTo-4 2000 20175948 ns/op 1.49 MB/s 175 B/op 3 allocs/op
PASS
```
Running this on a RPi3 gives the following:
```
$ go test -run ^$ -bench . -benchmem -benchtime 30s
goos: linux
goarch: arm
BenchmarkRoundTrip-4 2000 20178499 ns/op 1.49 MB/s 174 B/op 3 allocs/op
PASS
```
Dan Kortschak