on types where the idea of length is not really applicable

to use from the debugger

so we can address more types of functions in the future, including,
but perhaps not limited to, byte-compiled functions

as for nil and t

printer code to print everything to a strbuf first, then to a port,
although old code not yet removed

eval only and not elsewhere; listp; and; or; not; resource measurements
counts objects; included register_object in new_object, as it is
called only from there anyway

including comparisons (even divide and modulo, because we can);
timing and resource measuring primitives; funcall; apply; list;
let; letrec

builtin *and* form object, special and normal. Various consequenting
reconstructions all over the place. Lambda, function, fset, defun now all
call make_function to create a function object that is already verified
as a function. This means, hopefully, that we don't need to verify any
function (and be it a constructed-on-the-fly lambda or special form)
on apply or evalfun.

builtins and lambdas and mus), which still did not help the mapcar;
made the function function a special form, which magically did

getting the symbol value as a function, which still did not help the
macpar function; instead implemented the "function" function (although
wronly, only on symbols, yet) and the associated #' reader macro -- and
a general mechanism to support other reader macros with the #.

stacktrace, and in the repl print only the returned (error) object;
output/error ports everywhere except in the port handling itself,
where it is always stderr

line number, column, pushback token, etc. that is passed through
everything as a parameter, so we can now (in theory) have multiple
readers at a time; fixed the reader SEGV bug (list initialization
in read_loop was null instead of nil); new xstrdup()

made load_file use the repl; fixed a reader bug (newline had the wrong
charclass, so the reader never came out of a comment); tried to fix
another reader bug that causes SIGSEGVs, but failed so far

files