Auto merge of #8175 - benschulz:set-timeout-panic, r=jdm

Fixed panic in ActiveTimers.set_timeout_or_interval.

`ActiveTimers.set_timeout_or_interval` asserts that the pipeline is not currently frozen. Apparently that is too strict. When pending network requests complete after a pipeline is frozen, scripts may be executed and a timer scheduled.

With these changes scheduling a timer while the pipeline is frozen behaves as if the timer was scheduled at the time the pipeline was frozen.

To reproduce the panic
 1. `./mach run -r http://google.com`,
 2. immediately click on any link and
 3. wait for the panic.

<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/8175)
<!-- Reviewable:end -->

components/script/timers.rs

@@ -139,8 +139,6 @@ impl ActiveTimers {
                                is_interval: IsInterval,
                                source: TimerSource)
                                -> i32 {
-        assert!(self.suspended_since.get().is_none());
-
         // step 3
         let TimerHandle(new_handle) = self.next_timer_handle.get();
         self.next_timer_handle.set(TimerHandle(new_handle + 1));
@@ -283,7 +281,10 @@ impl ActiveTimers {
     }
 
     fn schedule_timer_call(&self) {
-        assert!(self.suspended_since.get().is_none());
+        if self.suspended_since.get().is_some() {
+            // The timer will be scheduled when the pipeline is thawed.
+            return;
+        }
 
         let timers = self.timers.borrow();
 
@@ -318,7 +319,12 @@ impl ActiveTimers {
     }
 
     fn base_time(&self) -> MsDuration {
-        precise_time_ms() - self.suspension_offset.get()
+        let offset = self.suspension_offset.get();
+
+        match self.suspended_since.get() {
+            Some(time) => time - offset,
+            None => precise_time_ms() - offset,
+        }
     }
 
     // see step 7 of https://html.spec.whatwg.org/multipage/#timer-initialisation-steps