getInputChar doesn't handle Unicode

[judah]

Currently 'getInputChar' doesn't handle Unicode, since utf8-string does not provide a unicode-aware 'getChar' equivalent.

This issue will probably be fixed once ghc provides unicode support for text I/O in 6.12.

[judah]

It's annoying, but we should be able to get most of the way without support from ghc:

• use iconv's EINVAL (incomplete input) to decide whether more characters need to be read.
• use Win32 API functions to correctly translate non-ASCII, single-byte characters from the code page to Unicode.
• use Win32's IsDBCSLeadByteEx to determine whether a second character needs to be read. (Note this won't work for UTF-8 on Windows; if someone complains, we could special-case UTF-8 or allow them to use iconv instead.)