Dynatrace file-based configuration slow on macOS

[pirgeo]

Thanks to @jonatan-ivanov for pointing out a problem with the WatchService based approach on macOS. A native implementation seems to be missing for the macOS file systems (at least in some JDKs), leading to a fallback to a polling implementation that checks the filesystem for changes every 10s. This is consistent with what @jonatan-ivanov reported in #2916. We are currently looking into this.

Original comment for context:

@pirgeo I think there are some issues with the Dynatrace file poller, it seems it takes a lot of time on macOS for the file poller to detect the changes.

The tests in this PR were failing for me on my mac (OS X 12.1, x86_64): see this report.
I replaced Thread::sleep with awaitility it takes almost 10 seconds on my machine to see the file changes (instead of 10ms that you had). Could you please look into it if anyone else has this issue before?

My changes: jonatan-ivanov@f4c1c42 (they are unrelated since the build was failing for me at the first place).

Originally posted by @jonatan-ivanov in #2916 (comment)