Log when pods are replaced and use null for LOG_HOME when effective log home is not set

[rjeberhard]

No description provided.

[moreaut]

I manually tested this against a previously built image containing a domain home and it worked for me.
Initially, I configured my domain to start 2 managed servers.
Then I changed the replica count to 1, and one ms terminated.
Then I changed the replica count back to 2, and one ms started.
The other running servers did not restart (as expected).