Add histograms to measure DNS lookups for different configurations.

Add histograms to measure DNS lookups for different configurations.
When we request the address family IPV4, we're asking for A records
only.  When we ask for family UNSPEC, we're asking for both A and AAAA
records.  Let's see if there is a latency difference.

BUG=none
TEST=n/a
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=88675

[jar (doing other things), 2011-06-09 18:30:00]

LGTM

Optional nit to consider below.

http://codereview.chromium.org/7003090/diff/3001/net/base/host_resolver_impl.cc
File net/base/host_resolver_impl.cc (right):

http://codereview.chromium.org/7003090/diff/3001/net/base/host_resolver_impl....
net/base/host_resolver_impl.cc:675: break;
nit: perchance add a default: NOT_REACHED.

The style guide suggests this for non-enumerated cases, but also claims that in
enumerated cases the compiler will warn.... but I doubt our compiler provides
such a warning.

[cbentzel, 2011-06-09 18:47:03]

Driveby comment-on-comment [this LGTM if it matters]

http://codereview.chromium.org/7003090/diff/3001/net/base/host_resolver_impl.cc
File net/base/host_resolver_impl.cc (right):

http://codereview.chromium.org/7003090/diff/3001/net/base/host_resolver_impl....
net/base/host_resolver_impl.cc:675: break;
On 2011/06/09 18:30:00, jar wrote:
> nit: perchance add a default: NOT_REACHED.
> 
> The style guide suggests this for non-enumerated cases, but also claims that
in
> enumerated cases the compiler will warn.... but I doubt our compiler provides
> such a warning.

I've had compiler warnings for that case [I believe gcc on Linux does, perhaps
others].