fix: Use PerformanceObserver to act as a second check to prevent returning 0 for duration and loadEventEnd

[adebayor123]

Version 1.1.0 and later has a bug which causes com.amazon.rum.performance_navigation_event.duration to be 0 when the web client loads before window.load has been fired.

Root Cause:

• Window fires load when the page has finished loading according to documentations, but there seems to be cases where it is actually fired before the loadEventEnd is fired.
• PR fix: capture PerformanceNavigationTiming events even when window.load fires before plugin loads #81 correctly fixes the bug that did not capture when the page loads before the plugin does, but by removing PerformanceObserver, we started emitting navigation_events as soon as window fired load.
• As a result, there are times when navigation_events emits 0 for fields such as duration, loadEventEnd, etc.

This PR addresses the issue by doing the following:

1. No longer rely on callback when the page has already loaded (we determine by using loadEventEnd) and directly record the values.
2. Restore PerformanceObserver for regular scenarios where page loads after the plugin.

Verifications:

1. Unit Test
2. Integ Test
3. https://paste.amazon.com/show/renoseo/1645745904

---

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[qhanam]

nit: The method's javadoc comment should explain the behavior of the method and optionally its inputs and outputs. Place implementation details as block comments inside the method.