Added file 'apply-substitition.lisp' with a few new functions that are
a start for the variable substitition operation.

New fixes to the MATCH and MATCH-CASE macros. They should now work as
advertised.

Minor changes to other files: added exports to package file, fixed
.system and .asd files.

Fixed two problems with the unifier machinery.
The first one had to do with the matching of NIL against SYMBOL and LIST
in several places: essentially, the problem is incongruencies in the
results of COMPUTE-APPLICABLE-METHODS in these cases. I think I caught
most of them: unification of lists and the occur-check were the obvious
places where things went awry.

The second problem had to do with the reader macro #T. The original
code generated an object at read time, which is not such a good idea.
Now the code generates a call to MAKE-TEMPLATE with is evaluated later.
Incidentally, the reader macro function is now called |sharp-T-reader|, in
order to placate Emacs fontification.

Modified Files:
templates-hierarchy.lisp unifier.lisp

Fixed problem with the unification of a list with a
SEQUENCE-TEMPLATE. The implementation was not checking that the
length of the list was compatible with the length of the required
elements in the template.

Apart from that, keyword matching is still unimplemented.

Fixed problem with checking the presence of T and OTHERWISE clauses in
MATCHING.

Fixed a documentation bug.
Thanks to Norman Werner for spotting it.

Fixed two major bugs reported by Norman Werver. Unification of
strings and symbols and of numers and symbols was not recurring on
VAR-UNIFY, as required; thus

(unify '(?x ?x) '("asd" "qweert"))
and
(unify '(foo ?x baz) '(foo 42 ?x))

would succed.
The two cases are now fixed.

Changed the DISCLAIMER parts in the files, in order to clarify the licensing
of the code.

Added IGNORABLE declaration to MATCH macro expansion.
This is useful in quieting various compilers.

Fixed bugs in COLLECT-TEMPLATE-VARS for NUMBER-TEMPLATEs.
If a variable was present, then the method was not returning a list.
The treatment of constants like PI was also incorrect, as the numeric
value was returned. Retunrning () seems the right thing to do
instead.

Added MATCH-CASE macro. Slightly modified from the version provided
by Peter Scott.

Fixed minor problem in Lambda List parsing.
Also VALID-TEMPLATE-P was expanded, and it looks like it should become
a generic function.