Add basic support for rational and complex numbers

[huonw]

This adds two generic data types, Ratio and Cmplx (and some aliases for useful instances, e.g. Ratio<int> and Cmplx<f64>), and basic arithmetic support, as well as .to_str (for both) and .from_str (for rational).

The complex number implementation doesn't solve #1284 other than getting something into the libraries, specifically it doesn't even try to address C interop. If the complex part of this gets merged, maybe it's worth closing that issue and reopening more specific issue(s) about the failings.

The implementations can be fleshed out when the numeric traits stabilise (and trait inheritance works).

[thestinger]

Two more things that can be turned into issues and dealt with in the future:

• Guarantee that the complex number type is compatible with the C representation #5737: It would be convenient to have the representation of complex numbers be compatible with C by making it packed, but I don't think that's currently possible. Luckily that can be changed later without even touching the interface at all.
• Figure out a good convention for normalizing rational numbers #5738: There's also the question about how frequently the rational number should normalize - Mpq does it after each operation, and then exposes a raw interface for doing a few operations without normalizing. If it's always guaranteed to be normalized, #[deriving(Eq)] might as well be used. That would probably also mean making the fields priv and using scary names for raw manipulation methods.