Fix backend issues and reset client db

[kchilleri]

This PR fixes issue #974.

client_db stores the following variables:
hostname of the front end/back end beeflow was started from, default is empty
backend flag: default is empty, "true" if beeflow was started on a compute node

When doing beeflow core start: client database is populated (--backend, when appropriate).
When doing beeflow core stop: client database gets reset, back to default values.

The issues appeared when beeflow core stop was not used and therefore the client database was not reset. I have added some checks for all cases I could think of when this happens. For example, If beeflow is started on a compute node and the allocation ends, or is relinquished, I have added checks every time beeflow core start is returned that check the database values and resets them accordingly.

Also, I added an error check that checks if the user is using the --backend flag when not on a backend node.

A test for the client database has been updated as well.