TimeDelta::FromDouble now uses saturated_cast

This change resolves the issue described in crbug.com/612601 (internal)
by changing the existing ternary implementation of TimeDelta::FromDouble
into a call to saturated_cast.

Back in 2016, when TimeDelta::FromDouble's current control flow
was written (https://crrev.com/1976703005), there was some concern
about using saturated_cast here because TimeDelta's permitted values
ranged from Max() to -Max() at the time. jyasskin implemented
a safe version [1] that met the [-Max(), Max()] requirement, but it
looks like the CL got lost somewhere before landing.

As of February 2018 (crrev.com/c/539058), the internal
value of TimeDelta can now range over all int64_t's, so the desired
semantics of TimeDelta::FromDouble are now those of
saturated_cast<int64_t>.

[1]
https://codereview.chromium.org/1976703005/diff/100001/base/time/time.h#newcode627

Test: Expanded time_unittest
Bug: 612601
Change-Id: Id00bdad9e732a22f0765c228b65ac6d65aa8b3d4
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1830114Reviewed-by: Asanka Herath <asanka@chromium.org>
Reviewed-by: danakj <danakj@chromium.org>
Commit-Queue: Asanka Herath <asanka@chromium.org>
Commit-Queue: David Van Cleve <davidvc@chromium.org>
Auto-Submit: David Van Cleve <davidvc@chromium.org>
Cr-Commit-Position: refs/heads/master@{#701704}

base/time/time.h

@@ -837,13 +837,7 @@ constexpr TimeDelta TimeDelta::Min() {
 
 // static
 constexpr TimeDelta TimeDelta::FromDouble(double value) {
-  // TODO(crbug.com/612601): Use saturated_cast<int64_t>(value) once we sort out
-  // the Min() behavior.
-  return value > std::numeric_limits<int64_t>::max()
-             ? Max()
-             : value < std::numeric_limits<int64_t>::min()
-                   ? Min()
-                   : TimeDelta(static_cast<int64_t>(value));
+  return TimeDelta(saturated_cast<int64_t>(value));
 }
 
 // static

base/time/time_unittest.cc

@@ -1352,6 +1352,10 @@ TEST(TimeDelta, MaxConversions) {
           .is_max(),
       "");
 
+  static_assert(
+      TimeDelta::FromMicrosecondsD(max_d).is_max(),
+      "Make sure that 2^63 correctly gets clamped to `max` (crbug.com/612601)");
+
   // Floating point arithmetic resulting in infinity isn't constexpr in C++14.
   EXPECT_TRUE(
       TimeDelta::FromMillisecondsD(std::numeric_limits<double>::infinity())