Move writability check out of `canUseUtf8` or drop it

[jeltsch]

#265 included a writability check to RIO.Prelude.Logger.canUseUtf8, because the previous implementation of canUseUtf8 also had one, although not so explicit, being covered by wantWritableHandle. However, this writability check does not belong in canUseUtf8, as it has nothing to do with encodings.

We should possibly move the writability check further up so that it is directly done during logger initialization.

An alternative is to simply drop the check and rely on actual writes failing when the handle is not writable. The state of a handle can be changed any time no operation is performed on the handle; in particular, the handle can be closed. Therefore, a check for writability during logger initialization anyhow provides no guarantee that the handle will be writable later, except perhaps if the handle is under the sole control of the logger.

[akhra]

no guarantee it's good later, but a significant debugging aid if it's bad already. it's an atrocious search space, handles can even potentially be closed outside the program, so every ruled-out possibility is valuable IMO.

100% in favor of moving the check, though! elsewhere I suggested we might want to think about refining the exception, but on further consideration that can be done later and this is a clear improvement regardless.

let's land #265, then I can take care of this one myself if you'd like to move on.

[jeltsch]

Yes, I‘d be happy if you’d take care of this. 🙂

[akhra]

addressed via #270, we'll want to move that into CPP once the upstream fix lands in a release.