Implement user-defined error handler mechanism

We could provide an option to let users define an error handler function which will handle all kinds of task failures (which may necessitate propagating more richly-annotated information on error source/reason). If this could be defined as a regular scheduler option (e.g. settable with `with_options`), it might alleviate questions about retries for failed operations.

This should also not be specific to errors from user task code - this should generalize to errors coming from basically anywhere in Dagger, including the storage system in #289 , the fault handler, etc.