Improve the hashtable related code by checking the node count

[ChengJin01]

The issue was detected when adjusting the thunk heap related code via eclipse-openj9/openj9#19852, in which we found that the code in hashTableForEachDo() still relies on the node count to determine whether to move forward to check the hashtable but there are still a bunch of code/assignments before returning null for a empty hashtable. Instead, we should check numberOfNodes early, which would be an improvement to hashTableStartDo(), hashTableForEachDo(), and possibly other hash table functions.

FYI: @keithc-ca