By keeping the repository url in the `packages` file scripts such as
source:ghc/sync-all can pick it up. Moreover, it's easier to
automatically validate for correctness than the
wiki:Repositories/Upstream page.

It's now always 'git'.

This makes the build system a little simpler, and in particular
will make it easier to handle the changes needed for cross-compilation.

This bring the time package into line with how we handle other
packages.

The *predicates* all start with "PKGS_THAT_...", e.g.:

  PKGS_THAT_BUILD_WITH_STAGE0   (previously "PACKAGES_STAGE0")
  PKGS_THAT_BUILD_WITH_STAGE2   (previously "STAGE2_PACKAGES")
  PKGS_THAT_USE_TH              (previously "TH_PACKAGES)
  etc. (there are a few more)

the lists of packages to build are now consistently named:

  PACKAGES_STAGE0
  PACKAGES_STAGE1  (previously just "PACKAGES")
  PACKAGES_STAGE2

It now matches PACKAGES_STAGE2

We no longer pull directly from upstream repos, so it wasn't useful
to have it programatically available. The info, and more, is now on
    http://hackage.haskell.org/trac/ghc/wiki/Commentary/Libraries#Repositorylocations
instead.

There is no longer any need for them to be separate, and this
way we don't have to call perl from the boot script. This is
particularly useful when perl isn't where we expect it to be
(#5154).

If we find one then we print an error message and fail.

If we aren't validating and mk/build.mk doesn't exist then boot
prints a warning, pointing at instructions for setting up mk/build.mk.

(which is the case in sdists)

It makes the parsing simpler if we always have the same number of columns

This stops us having to worry about sh/sed/... portability.

Now that the clean rules don't require libraries/*/ghc.mk, we don't
have to build them on demand.  And having them built on demand
introduced a failure mode (where some libraries have ghc.mk and some
don't).

An incorrect libraries/ prefix was being added.

The packages file has an extra component (for "darcs" or "git") so we need
to take that into account when grepping.

find on Windows doesn't understand -L, so stop trying to be clever and
just autoreconf everything.

Also, print out the names of directories as we autoreconf them, so that
if autoreconfing one breaks then we know which one it was.

Two problems here: find needs to dereference symbolic links (-L
option, I really hope that's portable), and we need to notice when
aclocal.m4 is updated.  

Somehow I think this was easier when it just always ran
autoreconf... what was wrong with that?

* Call autoreconf only if necessary, avoiding a strict tool dependency.
* Be a good *nix citizen and don't clutter the output when everything is fine.
* Make a few more scripts executable.