Reuse same browser for all pages in rasterize_and_record_micro

Reuse same browser for all pages in rasterize_and_record_micro

Don't restart the browser between each tested page in
rasterize_and_record_micro. This improves stability and -- somewhat
surprisingly -- reduces noise in the results. As an added bonus the
rasterize_and_record_micro.key_silk_cases test completes in about 9 min
compared to an earlier 16 min.

To measure the impact of the change I ran 18 rasterize_and_record_micro
iterations for 19 pages in key_silk_cases on a Nexus 5 (strapped to a
water bottle for cooling) and logged the per-page record_time results
for each run. The results were:

With browser restarts (baseline):
- Average record time: 1.78 ms
- Average record time standard deviation: 0.0978

Without browser restarts (patch):
- Average record time: 1.63 ms
- Average record time standard deviation: 0.0379

BUG=355952
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=273984

[tonyg, 2014-05-30 16:36:30]

lgtm

Woot! Are there other benchmarks that could benefit from the same?

[tonyg, 2014-05-30 16:36:47]

The CQ bit was checked by tonyg@chromium.org

[commit-bot: I haz the power, 2014-05-30 16:39:26]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/skyostil@chromium.org/306923007/1

[ernstm, 2014-05-30 16:41:53]

On 2014/05/30 16:36:47, tonyg wrote:
> The CQ bit was checked by mailto:tonyg@chromium.org

That is indeed surprising. In the past, the standard deviation was higher
without browser restart and I recently saw it being higher with repaint
benchmark as well.

[Sami, 2014-05-30 16:41:57]

Grep only tells me about 3 tests that do this: GPU context lost test, GPU
process test and startup perf tests. There could be others which evade grepping
like the one in this patch.

[commit-bot: I haz the power, 2014-05-31 01:39:30]

Change committed as 273984