Fix race/error handling issues on NBLS restart.

[sdedic]

During testing, it turned out that when NBLS is enabled "fast enough" after extension start, the extension may not initialize properly.
There are surely more bugs ;)) but the ones most prominent are:

• there was a missing catch() after stopClient() call
• reject() on kill promise was called before kill actually completed
• testAdapter was not re-initialized (and re-loaded with data) each time client restarts

In addition, if the client is stop()ped during/soon after restart, the Explorer (project view, database view or cloud assets) attemp to communicate with a dead client, which results in error notifications.

[sdedic]

During further testing it occurred that the request to restart NBLS because of change enableJava flag was fired before the logs were received and activationPending was re-set .... so the request to restart NBLS was ignored. It seems that the ancient "collision avoidance" is now fully obsolete by waiting on the client promise for stop().
I've removed the simple avoidance flag - but needed to slightly change code so that closed server will not respawn itself if another one is already starting.

Please review carefully.

[sdedic]

Squashed before merge