Makes it more consistent, imo.

They're still parsed for bootstrapping purposes, but the qualifier is immediately dropped on the floor.

r? @nikomatsakis 

They're still parsed though, to get through bootstrapping.

At the moment this only includes type checking and there is no code generation support yet. I wanted to get the design reviewed first.

From discussion with @graydon at #5841, re-implemented as `#[simd]` attribute on structs.

Progressing towards #3499.

r? @ILyoan

This pulls all the logic for discovering the crate entry point into a new pass (out of resolve and typeck), then changes it so that main is only looked for at the crate level (`#[main]` can still be used anywhere).

I don't understand the special android logic here and worry that I may have broken it.

I don't understand what this logic is doing

There's no need to delegate to C to call the Rust main function.

This "finishes" the generic deriving code (which I started in #5640), in the sense it supports everything that I can think of being useful. (Including lifetimes and type parameters on methods and traits, arguments and return values of (almost) any type, static methods.)

It closes #6149, but met with #6257, so the following doesn't work:
```rust
#[deriving(TotalEq)]
struct Foo<'self>(&'self int);
```
(It only fails for `TotalOrd`, `TotalEq` and `Clone`, since they are the only ones that call a method directly on sub-elements of the type, which means that the auto-deref interferes with the pointer.)

It also makes `Rand` (chooses a random variant, fills the fields with random values, including recursively for recursive types) and `ToStr` (`x.to_str()` is the same as `fmt!("%?", x)`) derivable, as well as converting IterBytes to the generic code (which made the code 2.5x shorter, more robust and added support for tuple structs).

({En,De}codable are trickier, so I'll convert them over later.)

#6312