Add SBClientDownload.DownloadRequestNetworkDuration and DownloadRequestTimeoutDuration histograms.

tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index 916791252a78aab5aef79c2dac1a06fdd6ef2d5f..3455b450bd3ee77227499ec35fe8d78f80a32320 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -23437,8 +23437,11 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   <owner>mattm@chromium.org</owner>
   <summary>
     Records the total time it takes for the SafeBrowsing download service to
-    check whether the content of a download is malicious or not. This histogram
-    only includes requests that are sent to the SafeBrowsing server.
+    check whether the content of a download is malicious or not, including file
+    feature extraction, whitelist checking, and server ping. This histogram only
+    includes checks that sent a ping to the SafeBrowsing server. It does not
+    include requests that were cancelled, but does include requests that
+    received a bad response.
   </summary>
 </histogram>
 
@@ -23449,6 +23452,16 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="SBClientDownload.DownloadRequestNetworkDuration"
+    units="milliseconds">
+  <owner>mattm@chromium.org</owner>
+  <summary>
+    Records the time it takes for the SafeBrowsing download service ping. It
+    includes both cancelled and completed requests. Cancelled requests will be
+    recorded as the time the request was cancelled.

[Mark P, 2014/06/04 16:47:34]

I would think you want two histograms here, one with two buckets counting the
numbers of successfully completed requests versus cancelled requests, and
another that shows the time distribution for successful requests.  (I assume the
time distribution for canceled requests is going pretty much to be a peak at the
value of a timeout in your code.  If not, then perhaps another histogram showing
the distribution of times for canceled requests.)  Right now you're adding two
separate distribution together and reporting them as one, and I'm not sure how
much insight this will get you.

[mattm, 2014/06/04 23:55:47]

Makes sense.  Did that for both the Network and Timeout histograms, hopefully
this isn't too much histogram spam.
Yeah, aside from cancellation due to shutdown which should be pretty rare.

+  </summary>
+</histogram>
+
 <histogram name="SBClientDownload.DownloadRequestPayloadSize" units="bytes">
   <owner>mattm@chromium.org</owner>
   <summary>
@@ -23464,6 +23477,16 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="SBClientDownload.DownloadRequestTimeoutDuration"
+    units="milliseconds">
+  <owner>mattm@chromium.org</owner>
+  <summary>
+    Records the portion of the SafeBrowsing download service check starting with
+    the point CheckClientDownloadRequest::StartTimeout() is called. It is
+    recorded regardless if a ping was sent or not.
+  </summary>
+</histogram>
+
 <histogram name="SBClientDownload.ExtractImageHeadersTime" units="milliseconds">
   <owner>grt@chromium.org</owner>
   <summary>