Re-land issue 750183008 (Parallelize trace messages serialization)

Re-land issue 750183008 (Parallelize trace messages serialization)

This re-lands https://codereview.chromium.org/750183008 (originally
by loislo@) with the output chunk size reduced down from 10M to 100K
(close to what it used to be originally). The reason for telemetry
failures of original change on Windows was the bug in websocket-client
(https://github.com/liris/websocket-client/issues/163) that caused
it to fragment Python heap when receiving a large frame.

Original issue description:
Move serialization into a worker thread. As a result
IO thread will be able to send messages to the browser.

The original implementation did serialization on IO thread
and was not able to send the messages because ipc had
is_blocked_on_write_ = true and had no chance to check
the actual state of the channel. So the messages were
collected in output_queue. Also the messages could be quite
big and could block the IO thread for a long time.

BUG=463572
TBR=dsinclair
(as he already l-g-t-m-d the original patch)

Committed: https://crrev.com/61b95998e904561aa31aa095982529543f2cbac5
Cr-Commit-Position: refs/heads/master@{#322391}

[caseq, 2015-03-25 17:28:29]

caseq@chromium.org changed reviewers:
+ dsinclair@chromium.org, loislo@chromium.org

[loislo, 2015-03-26 14:57:48]

lgtm

[caseq, 2015-03-26 14:59:04]

caseq@chromium.org changed reviewers:
- dsinclair@chromium.org

[caseq, 2015-03-26 14:59:08]

The CQ bit was checked by caseq@chromium.org

[commit-bot: I haz the power, 2015-03-26 14:59:45]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1035783002/1

[commit-bot: I haz the power, 2015-03-26 15:56:51]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-03-26 15:57:25]

Patchset 1 (id:??) landed as
https://crrev.com/61b95998e904561aa31aa095982529543f2cbac5
Cr-Commit-Position: refs/heads/master@{#322391}