Consider adding more comprehensive protection against use-after-suspend

[Aaron1011]

Currently, we prevent users from trying to produce multiple ControlFlow::Suspend errors within a single execution run of a durable task (e.g. calling ctx.sleep_for multiple times without awaiting and propagating ControlFlow::Suspend).

We should consider doing more:

• Block all usage of the TaskContext after ControlFlow::Suspend is produced (e.g. don't allow spawning any new tools).
• Adjust the durable SQL functions to prevent the 'wait' state from getting overwritten for a particular task run (though we need to make sure that this doesn't break retries)