CL 136796 introduced benchmarks and refactored tests to use a
common list of test images. The tests now fail when run with
count > 2 since they rely on a fresh image each run.

Fix this by changing the list of test images to a list of test
image generator functions.

Change-Id: I5884c6bccba5e29bf84ee546fa501bc258379f42
Reviewed-on: https://go-review.googlesource.com/137295Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>