Restored change from error to warning from CL 85959

[eernstg]

The change from 'compile-time error' to 'static warning' that we performed in CL 85959 was accidentally lost when 'invalid_returns.md' was integrated into the language specification. This CL just changes those few words back to 'static warning' (and removes a couple of spurious {}).

[lrhn]

I still think this should be an error.
@leafpetersen

[eernstg]

We could do that:

Given that tools have reported an error for this situation for a while (we currently get expectation 'StaticWarning' and actual outcome 'CompileTimeError' for, e.g., analyzer-linux-release language_2/void/return_future_future_or_void_async_error1_test/none), we are in a good position to just keep the specification unchanged (so it remains an error, now per decision rather than by accident), that is: we'd just delete this pull request.

We might want to consider all the other warnings an extra time (maybe some of them can also be errors), but in that case we'd need to change the specification, so that would be done in a separate PR.

[leafpetersen]

(we currently get expectation 'StaticWarning' and actual outcome 'CompileTimeError' for, e.g., analyzer-linux-release language_2/void/return_future_future_or_void_async_error1_test/none)

This seems to have been an (accidental?) regression between 2.0 (where the analyzer reports a warning) and 2.1 (where it is an error). Maybe reasonable to move forward with it since it's been this way for a while? Not great that this slipped through though.

[eernstg]

Abandoned. At this point we're more likely to change even more warnings to errors, rather than changing these two errors to warnings.