Switched from loading test files to testing specific parsing constructs. The tests don't fully cover the span (eventually I'll add more, maybe), but there's a bunch there.
More to the point, the yaml-reader program doesn't do any file IO, so it doesn't require a specific working directory.

It's better organized now, I think - nodes only offer a single main way of getting the fundamental scalar (as a string), and now we can specialize a single template to read specific types.

In particular:
1. Windows CR/LF weren't read properly (issue #11)
2. Scanning wasn't reading EOF properly
3. Documents may be empty (this was old, I think)
Also fixed some VS2008 warnings on /W4.

Switched to reading the entire file into a buffer at the start.\nThis speeds it up a TON (like 100x).

Re-added the throw() specification to ~Exception(), and also to ~TypedKeyNotFound(); I suppose this'll fix the gcc compiler error.

Removed throw() specifier in Exception (I don't remember putting it in). This may solve a gcc error (I haven't tested it yet) or it may break it further.

Added a templated derived exception to KeyNotFound so that you can figure out *which* key wasn't found.

(Thanks to Vadim Zeitlin)

Added Read() functions for Node that return true/false, so we can easily check if a read is successful without throwing.
But we still have operator >> that throws on failure.