Fix windows packaging tests

[alpar-t]

On windows, it happens that the process we called terminates but some
other process it creates still has the same output strems and thus the
files open, so we can't clean it up.

This PR makes the cleanup a best effort.

[elasticmachine]

Pinging @elastic/es-core-infra (:Core/Infra/Packaging)

[rjernst]

This seems extreme to do for all commands we run on windows within packaging tests. If there is a child process created by some command, why are we not waiting on that child?

[alpar-t]

@rjernst AFAIK this happens when we run elasticsearch -d so we don't actually want to wait. This caused a lot of problems as it was previously causing the test to hang.
If it's ok with you I would like to look at this in a follow-up. At least we'll have the logs to see where it's happening.
My goal is to get the tests passing to be able to add PR checks to make sure the proper care will be taken not to break them.
That will make exploring additional changes easier too, right now testing everything before opening a PR is manual and takes a lot of time.

[alpar-t]

I'm going to merge this to prevent CI failures and hopefully get some successful runs over the weekend. Will act on any feedback following that.