Split default encoding into file system encoding and content encoding

We have a default text encoding setting right now which is used as a fallback when we need to pre-configure the encoding or when we cannot determine the best encoding to use. This is currently defaulted to utf-8.

This works fine for file content, but there is a problem with extended characters on the file system, which are usually reported in the user's own codepage as defined by their regional settings. Ideally when we're calling commands which return or process file names, we should use the default file system code page that's in use. This means having 2 different default text encodings, one for text content and one for file names. The former should still be defaulted to utf-8, and the latter should be defaulted based on the user's OS regional settings.