Improve reliability of File Watchers in Linux containers

This issue is related to #5564

If you create a function on a new dedicated Linux app in the portal, file watchers won't reliably pick up the changes in order to restart the ScriptHost. This is apparently a known issue with file watchers in linux containers, especially if monitoring a mounted drive.

One potential fix would be to use a polling file watcher instead. We'd probably want to do this only if it's linux because the windows file watchers seem to be working fine.

Start by reproducing the unreliable restart behavior when a function is created or changed in a Linux container using a mounted drive. Investigate the host's file-watcher entry points and compare Linux behavior with Windows; done means changes are reliably detected and the ScriptHost restarts without regressing Windows watchers.

Written by the indexing model from the issue text.