Added check in OnChannelError to see if metro/ash desktop is created.

Added check in OnChannelError to see if metro/ash desktop is created.
As this bug is not repro in lab environment or locally and has been there for very long time, imagined and exprimented locally, where I intentionally crashed metro viewer before sending SetTargetSurface back to browser after channel connection and could repro this issue. Cause of why viewer process fails is still unknown, but there could be number of reasons and needs to be addressed separately.

This bug seem to be assigned to various people with different issue numbers, I plan to combine them together.
Cause of latest crash uptick might be Avast (as it seems more common than others), but low number of ClientID's and low percentage for Avast dll in rest of Chrome versions doesn't lead to any concrete conclusion.

BUG=487803
R=cpu@chromium.org,ananta@chromium.org
TBR=sky@chromium.org

Committed: https://crrev.com/9c5a337c1943dc4f123025e0d9e5cb15ef4a0a3d
Cr-Commit-Position: refs/heads/master@{#330036}

[Shrikant Kelkar, 2015-05-14 22:43:11]

shrikant@chromium.org changed reviewers:
+ ananta@chromium.org

[cpu_(ooo_6.6-7.5), 2015-05-15 01:28:46]

lgtm

[Shrikant Kelkar, 2015-05-15 02:59:18]

The CQ bit was checked by shrikant@chromium.org

[commit-bot: I haz the power, 2015-05-15 02:59:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1128823009/1

[commit-bot: I haz the power, 2015-05-15 03:46:20]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-05-15 03:47:04]

Patchset 1 (id:??) landed as
https://crrev.com/9c5a337c1943dc4f123025e0d9e5cb15ef4a0a3d
Cr-Commit-Position: refs/heads/master@{#330036}