Document that getFileSystemEncoding can return ASCII

Summary

https://gitlab.haskell.org/ghc/ghc/-/blob/9fc0fe008c13782cb7b1962b0ebed0bb09ecfb6f/libraries/base/GHC/IO/Encoding.hs#L112-123 claims that GHC.IO.Encoding.getFileSystemEncoding returns "Unicode encoding of the current locale". It seems to be incorrect: on some systems it returns ASCII, which is not a Unicode encoding.

Steps to reproduce

On an Ubuntu host run a virtual s390x machine:

apt install -y docker.io && \
docker run --rm --privileged multiarch/qemu-user-static --reset -p yes && \
docker run --rm -i -t s390x/ubuntu

Then on the virtual machine:

apt update && \
apt install -y ghc && \
ghc --version && \
ghci

and finally in ghci on the virtual machine:

> enc <- GHC.IO.Encoding.getFileSystemEncoding
> enc
ASCII
> GHC.Foreign.newCStringLen enc "\160"
*** Exception: recoverEncode: invalid argument (invalid character)

Proposed changes

I'm filing this as a documentation issue, because foremost the comment for getFileSystemEncoding should not mislead developers that it returns Unicode encodings only. But it would also be nice to make the exception from newCStringLen more helpful, at least mentioning target encoding and ideally a stack trace as well. I hit the issue while running bytestring test suite on s390x, and as one can imagine debugging was not pleasant at all.

Environment

• GHC version: 8.6.5
• Platform: s390x