zebra's feedback tackled.

Mostly ACK, but i've noticed what looks to be a race condition with the directory locks not always being released before the restart happens. this behavior isn't introduced in this PR itself, but since we're on the subject of the restart process, this is as good a place to address it.

Simply adding a call to ReleaseDirectoryLocks() at the end of Shutdown() in init.cpp has show to clear up the issue for me

yeah, good catch.
I started doing this cleanup to investigate the race condition but stopped once i saw the ugly CExplicitNetCleanup::callCleanup function and all the other stuff that cleaned..
it's all related to the same cause: the restart functionality, as it was coded, starts a new process at the same time that the running process is being stopped, so there is a race condition over every shared resource (not only the dirs locks, there could be others).

Better to spend a bit more time looking for a cleaner way of doing the restart before moving forward with the introduction of another manual resource cleanup only for the restart flow and search if there is any other static field that is destructed/released at the program's exit that can be affected by this function's behavior. It smells bad.

rebased on master, conflicts solved. Ready.

The comment before Shutdown(), mentioning the skip of PrepareShutdown() in case of restarts is now outdated.
In fact, we don't even need two separated steps anymore (can pick here if you want random-zebra@030d037)

Side note: #2646 will be rebased (removing RestartRequested()) after this PR and #2641 are merged. We could include that one too for 5.4 as it's very simple.

yeah, added your commit zebra. Further cleanups are more than welcome 👌. This process was full of redundancies.

all good for me to add #2646 to v5.4 as well. Let's just do it after branching off and starting the public testing phase.