Disable logging of failed SPDY PINGs in a timemax bucket, and update histogram...

Disable logging of failed SPDY PINGs in a timemax bucket, and update histogram bucketing strategy for Net.SpdyPing.RTT to match the strategy of Net.TCP_Connection_Latency.
* timemax is int64max, whereas histograms use int values, so this doesn't end up working well and can break certain analysis operations such as computing the average SPDY PING time. Additionally, counts for failed SPDY PINGs are recorded in the Net.SpdySession.ClosedOnError enum histogram, so there's no need for us to log them in a special bucket in the RTT histo as well.
* Also change the histogram bucketing strategy for Net.SpdyPing.RTT to match the strategy of Net.TCP_Connection_Latency since these are measuring similar things and it's useful to be able to compare TCP RTTs with application-layer RTTs.

BUG=475129

Review URL: https://codereview.chromium.org/1074523002

Cr-Commit-Position: refs/heads/master@{#324318}

net/spdy/spdy_session.cc

@@ -2960,8 +2960,6 @@ void SpdySession::CheckPingStatus(base::TimeTicks last_check_time) {
   base::TimeDelta delay = hung_interval_ - (now - last_activity_time_);
 
   if (delay.InMilliseconds() < 0 || last_activity_time_ < last_check_time) {
-    // Track all failed PING messages in a separate bucket.
-    RecordPingRTTHistogram(base::TimeDelta::Max());
     DoDrainSession(ERR_SPDY_PING_FAILED, "Failed ping.");
     return;
   }
@@ -2975,7 +2973,9 @@ void SpdySession::CheckPingStatus(base::TimeTicks last_check_time) {
 }
 
 void SpdySession::RecordPingRTTHistogram(base::TimeDelta duration) {
-  UMA_HISTOGRAM_TIMES("Net.SpdyPing.RTT", duration);
+  UMA_HISTOGRAM_CUSTOM_TIMES("Net.SpdyPing.RTT", duration,
+                             base::TimeDelta::FromMilliseconds(1),
+                             base::TimeDelta::FromMinutes(10), 100);
 }
 
 void SpdySession::RecordProtocolErrorHistogram(