Previously an unimplemented error was thrown when using #[deriving_eq] on tuple-like struct definitions.

Rid libsyntax of records and get rid of the last piece in `librustc/front/test.rs`.

librustc: Separate the rest of the trait bounds with `+` and stop parsing space-separated ones. rs=plussing

...because it appears to work now. Removes a FIXME.

I've moved all intrinsics in a single file (libcore/private/intrinsics.rs) and changed a few files to make use of this file (e.g. vec.rs: move_val_init).

Two intrinsics have been commented out:
visit_tydesc: it uses TyDesc and TyVisitor, this would create a dependency on librustc which seems undesirable.
frame_address: I really had no idea what it should look like without the legacy modes (would it even work? In several places in libcore the (legacy-modes) intrinsics were wrapped
with a normal fn) and what it was supposed to do.

Some documentation is still required, but many names are fairly self-explanatory.

Most work done by @luqmana and @pcwalton - I just rebased.

1. Highlight new lifetime syntax
2. Visually distinguish `unsafe` keyword
3. Add a new file that highlights column 78, to warn when lines get too long

...because it appears to work now. Removes a FIXME.

r?

Previously check always succeeded because struct type was derived from the matched expression, not the matched pattern.

Fix #4849.

Changed a test to use intrinsics.rs and changed a few lines that were too long, had bad whitespace, etc.

rustc intrinsic. Several files in libcore have been changed to use
these intrinsics.
As of yet, none of the intrinsics are documented.

1. Highlight new lifetime syntax
2. Visually distinguish `unsafe` keyword
3. Add a new file that highlights column 78, to warn when lines get too long

Closes #4924

There are some more `Load` calls that could use this, so it's not completely done yet.

It looks like the type for the whole pattern wasn't being written back, so when privacy tried to look it up everything died.

Closes #4924

As far as I can tell, the only reason run-pass/type-use-i1-versus-i8
is trying to do a read is because that code was left over from the
original program the issue was found in.  When that test is run as
part of check-fast, and apparently only in that case, the test blocks
indefinitely, which is bad.