When using the shell for spawning the master and slave integration
processes the shell becomes the process that is returned by the spawn
command. This in turn means that the SIGINT that was supposed to be
sent to the master and slave processes got ignored by the shell instead.
Fix by spawning the node process directly instead of through a shell.
Shutdown running instances and close the connection to the master server
on slave stop. Remove call to process.exit on shutdown as it's no
longer needed.
Use SIGINT instead of the default SIGTERM when signalling to shut down
the integration master and slave subprocesses. This should let nyc
properly write the coverage report before it exits.
Spawn master, slave, and clusterctl in separate processes in the
clusterio integration tests and leave the test master and slave process
available for the whole test duration to allow plugins to use them as
well. This removes the need for reimplementing the startup logic in the
test code and makes the existing startup code covered by tests.
Hornwitser