Add E2E tests which verifies that Scheduler can clear offline peer metadata when the host exits normally/abnormally.

[BruceAko]

Feature request:

Earlier we have added functionality so that hosts send LeaveHost() requests to the Scheduler on normal exit and hosts on abnormal exit are also sensed by the Scheduler GC after a set amount of time and the metadata is cleared. Now, we are going to add E2E tests to verify that this functionality is correct.

Use case:

Verifies that Scheduler can clear offline peer metadata when the host exits normally/abnormally.

Steps:

1. Get the number of all hosts in the scheduler.
2. After a normal exit of a host, verify that the number of hosts in the scheduler decreases by 1.
3. After an abnormal exit of a host, verify that the number of hosts in the scheduler decreases by 1 after twice the announce host interval.