State of default constructed event should be re-considered

This issue raises a reasonable (IMHO) use case of events:
https://github.com/Apress/data-parallel-CPP/issues/20

Default constructed events initialize to the ready state in SYCL 2020:

> Constructs an event that is immediately ready

This seems like a confusing and error prone design point. Instead I'd expect the code in the linked issue to either work as the developer intended (wait and operate out of order) in the unsignaled interpretation, or to throw a synchronous exception that invalid events have been used on the initial `parallel_for` submission.

The current spec may make it easier to introduce program bugs than originally thought. Should be discussed within the Khronos WG.

Start by reading the SYCL 2020 wording that says a default-constructed event is immediately ready, then review the linked data-parallel-CPP example. Compare the proposed unsignaled interpretation and synchronous-exception behavior, and bring the alternatives to the Khronos WG; done requires a decided specification direction and corresponding wording.

Written by the indexing model from the issue text.