This large commit combines several interrelated changes:

  - IfaceSyn now contains actual Names rather than the special
    IfaceExtName type.  The binary interface file contains
    a symbol table of Names, where each entry is a (package,
    ModuleName, OccName) triple.  Names in the IfaceSyn point
    to entries in the symbol table.

    This reduces the size of interface files, which should
    hopefully improve performance (not measured yet).

    The toIfaceXXX functions now do not need to pass around
    a function from Name -> IfaceExtName, which makes that
    code simpler.

  - Names now do not point directly to their parents, and the
    nameParent operation has gone away.  It turned out to be hard to
    keep this information consistent in practice, and the parent info
    was only valid in some Names.  Instead we made the following
    changes:

    * ImportAvails contains a new field 
          imp_parent :: NameEnv AvailInfo
      which gives the family info for any Name in scope, and
      is used by the renamer when renaming export lists, amongst
      other things.  This info is thrown away after renaming.

    * The mi_ver_fn field of ModIface now maps to
      (OccName,Version) instead of just Version, where the
      OccName is the parent name.  This mapping is used when
      constructing the usage info for dependent modules.
      There may be entries in mi_ver_fn for things that are not in
      scope, whereas imp_parent only deals with in-scope things.

    * The md_exports field of ModDetails now contains
      [AvailInfo] rather than NameSet.  This gives us
      family info for the exported names of a module.

Also:

   - ifaceDeclSubBinders moved to IfaceSyn (seems like the
     right place for it).

   - heavily refactored renaming of import/export lists.

   - Unfortunately external core is now broken, as it relied on
     IfaceSyn.  It requires some attention.

See test rn003

Avoid possibility of name clash

Fixes Trac #925

A new comment in TcRnDriver in tcRnStmt reads thus: 

At one stage I removed any shadowed bindings from the type_env;
they are inaccessible but might, I suppose, cause a space leak if we leave them there.
However, with Template Haskell they aren't necessarily inaccessible.  Consider this
GHCi session
	 Prelude> let f n = n * 2 :: Int
	 Prelude> fName <- runQ [| f |]
	 Prelude> $(return $ AppE fName (LitE (IntegerL 7)))
	 14
	 Prelude> let f n = n * 3 :: Int
	 Prelude> $(return $ AppE fName (LitE (IntegerL 7)))
In the last line we use 'fName', which resolves to the *first* 'f'
in scope. If we delete it from the type env, GHCi crashes because
it doesn't expect that.

- Class and type family instances just got a lot more similar.
- FamInst, like Instance, now has a rough match signature.  The idea is the
  same: if the rough match doesn't match, there is no need to pull in the while
  tycon describing the instance (from a lazily read iface).
- IfaceFamInst changes in a similar way and the list of all IFaceFamInsts is
  now written into the binary iface (as for class instances), as deriving it
  from the tycon (as before) would render the whole rough matching useless.
- As a result of this, the plumbing of class instances and type instances 
  through the various environments, ModIface, ModGuts, and ModDetails is now
  almost the same.  (The remaining difference are mostly because the dfun of a
  class instance is an Id, but type instance refer to a TyCon, not an Id.)

*** WARNING: The interface file format changed! ***
***	     Rebuild from scratch.		***

Mention that the module sub-directory structure for .o and .hi files is created automatically by GHC

With the new Haddock patch, renamer errors can be duplicated;
so we want to bale out before doing the Haddock stuff if errors
are found.

(E.g test mod67 shows this up.)

This patch changes HscTypes.lookupIfaceByModule.  The problem was that
when compiling the 'base' package, we'd repeatedly reload GHC.Prim.
This is easily fixed by looking in the PIT too. A comment with
lookupIfaceByModule explains

This patch fixes Trac #909.  The problem was that when compiling 
the base package, the handling of wired-in things wasn't right;
in TcIface.tcWiredInTyCon it repeatedly loaded GHC.Base.hi into the
PIT, even though that was the very module it was compiling.

The main fix is by introducing TcIface.ifCheckWiredInThing.

But I did some minor refactoring as well.

This is a slight improvement over the patch sent by jamey@minilop.net,
we now do it properly if the source repo was a GHC tree on the local
filesystem too.

Merge post 6.6.

Comment Note [Take care] explains.

mkAtomicArgs is a mess.  A substantial rewrite of Simplify is needed.

This patch teaches SpecConstr about casts; see Note [SpecConstr for casts]

Note [Float coercions]
~~~~~~~~~~~~~~~~~~~~~~
When we find the binding
	x = e `cast` co
we'd like to transform it to
	x' = e
	x = x `cast` co		-- A trivial binding
There's a chance that e will be a constructor application or function, or something
like that, so moving the coerion to the usage site may well cancel the coersions
and lead to further optimisation.  Example:

     data family T a :: *
     data instance T Int = T Int

     foo :: Int -> Int -> Int
     foo m n = ...
        where
          x = T m
          go 0 = 0
          go n = case x of { T m -> go (n-m) }
		-- This case should optimise