Autodetect text/plain mimetype files instead of assuming utf-8

[thomasjohns]

Probably https://github.com/chardet/chardet or https://github.com/PyYoshi/cChardet could be used.