Assume UTF-8 encoding for .markdown and .haml - using new Git.readFile functionality

[pirxpilot]

Only pull after new version of git-fs is published to npm.
This should fix encoding problems #18, #28 etc.
Tested with howtonode.org - seems OK.

Reads .markdown, .haml and snippets as "utf8"
Everything else is read as raw buffers - no conversion needed.

[creationix]

Why not read everything as buffers. They automatically toString to utf8 strings when coerced.

[pirxpilot]

I had another look. I am pretty sure my patch loads them as strings (i.e. explicitly passing 'utf8' encoding) only when they are used as strings. The results of Git.readFile are passed to 'Haml.compile' and various string methods (substr, match etc.) are called on them: I don't think such usage triggers implicit coercion.

There was actually one explicit conversion that I was able to remove from prepareMarkdown:

if (!(typeof markdown === 'string')) {
  markdown = markdown.toString();
}

After reading your comment I experimented with loading everything as buffers and wheat predictably breaks in couple of places.
What am I missing?