Return NIL if the sequence is empty, instead of the NO-EXTREMUM nonsense.

 It was bad design, because it's not an error someone higher up the stack can
 sensibly handle, and handling it locally is too verbose and slow.

    (or (extremum ...) (error ...))

 expresses the common case succintly, and fits the pattern of existing
 sequence functions.

 If it is deemed necessary, we can also add &KEY DEFAULT, but that seems
 overkill and has little precedent in sequence functions.

  From http://www.cliki.net/EXTREMUM

  A simple version built on top of REDUCE for now.

  Yech. Phui. Tabs.

  Workarounds for the Clisp and CMUCL failures.

  Newer Clisp's no longer have the stack-overflow for the hash-table test, so
  enable it too.

  Added test-case.

  This reverts commit 485544d4.

  MEAN and MEDIAN are again non-generic, and DISPLACE-ARRAY is removed.

  Motivation on the mailing list.

 Objects other than sequences have means and medians (eg probability
 distributions, arrays, sequences/arrays wrapped in another object, etc), so
 it makes sense to make these functions generic.

 DISPLACE-ARRAY is a small utility function that is used in the implementation
 of MEDIAN for arrays, but is also of general utility because it makes the
 creation of displaced arrays simpler, dispensing with the need to specify the
 element type, and offering sensible defaults when one wants vectors.

 Also added tests for all the new functions/methods.

  Multiple evaluation of the function argument, oops.

  Plus tests.

  No need to depend on the vagaries of ADJUST-ARRAY.

  CIRCULAR-TREE-P had an unfortunate corner case, causing it to overflow
  the stack (seen and repoted at #lisp for '#1=(#1#).)

  The problem is caused by the end-test (of the outer DO) being run before the
  body has a first chance to check for (member slow seen).

 Patch by Gustavo on alexandria-devel.

 Also add tests.

Patch by: Peter Graves <peter@armedbear.org>

Added new test cases for DISJOIN, CONJOIN, COMPOSE and MULTIPLE-VALUE-COMPOSE.
All of them testing for the edge case of a single predicate / function.

- can't assume that (upgraded-array-element-type 'fixnum)

- COPY-HASH-TABLE.1 causes a stack overflow due to a CLISP bug.
- ALIST-HASH-TABLE.1, PLIST-HASH-TABLE.1 fail because HASH-TABLE-TEST
  returns EXT:FASTHASH-{EQ,EQL}.

- TYPEP returns a generalized boolean.  On Lisps like ECL, this detail
  was causing some tests to fail.  ERRORP takes this into account.

- make it usable for other Lisps besides SBCL.
- TEST-OP runs tests both compiled and evaluated.

- New function: PROPER-LIST-LENGTH.

- Share code between PROPER-LIST-LENGTH, LASTCAR, and SETF LASTCAR
  via a macrolet.

- Use ENDP wrapped in SAFETY 3, to ensure proper detection of dotted
  lists.

- Use PROPER-LIST-LENGTH instead of LIST-LENGTH in ROTATE, RANDOM-ELT,
  and SHUFFLE, and test that they signal an error for improper lists.
  (Thanks to Tobias Rittweiler)

 Editing damage: extra IF surrounding the permutation guts. Also add a
 trivial test-case.

	* conditions.lisp (unwind-protect-case): New macro. Similiar to
	CL:UNWIND-PROTECT except that it's possible to explicitly specify
	under which circumstances cleanup operations are run.

	* tests.lisp (unwind-protect-case.1-5): New test cases.

	* types.lisp: Fix quoting problem introduced by Attila's last
	patch; additionally add automatically generated docstrings to all
	the types defined.

	* tests.lisp (cdr5.*): New test cases.

	* hash-tables.lisp (copy-hash-table): Added new &key arg :KEY;
	it's run on each element before it's stored in the new HT.
	Additionally, make all &key args get their default even when
	NIL was passed for them.

	* tests.lisp (copy-hash-table.2): New test case.

	* tests.lisp (define-constant.1): Adapted to latest changes;
	argument to :KEY is now evaluated.

 * More efficient with constant arguments.

 * Respect lexical environment with non-constant arguments.

 * Support (once-only ((nx x)) ...) style also.

 * Like DOLIST, but iterates over plists.

 * ...so that running tests multiple times caused unrelated tests to fail
   due to coalesced constants being frobbed. Gah.

 * Thanks to Andreas Fuchs -- I only took the liberty of changing the name from
   MULTIPLE-VALUE-OR to NTH-VALUE-OR.

 * So SANS is now REMOVE-FROM-PLIST.

   ...I have to say that I'm still not sure I like this:

     (remove-from-plist x y) ; which is the plist?

   The common usage in REMOVE &co is to put the element designators
   first. This is confusing.

   Maybe we really want both:

    function SANS plist &rest keys
    function REMOVE-FROM-PLIST keys plist

 * plus a test-case