Fix deadlock between node removal and closing the app

[slawek-es]

There is a deadlock issue on App._dom_lock, where _close_all holds the lock and waits for the task associated with a message pump, but that task may be executing a "called-next" _prune_nodes call driven by an AwaitRemove which was triggered, for example, by the Footer.recompose call from handling the bindings_updated signal. This deadlock leads to a TimeoutError on app close.

We fix this by omitting the call to _prune_nodes in the AwaitRemove task if the whole application is already closing; it does not make sense to remove individual nodes in this case anyway.

Fixes #4677

Please review the following checklist.

• [x] Docstrings on all new or modified functions / classes
• [x] Updated documentation
• [x] Updated CHANGELOG.md (where appropriate)

[willmcgugan]

This will prevent Unmount messages from being sent.