[CM] More trouble with (values) from C with bonus segfault

[bil at ccrma.Stanford.EDU]

s7_define_function returns an unsafe function (there's 
s7_define_safe_function
for the safe case), but s7_define_typed_function returns a safe 
function.  This
inconsistency arose a long time ago when I noticed that all but a 
handful of
the functions in Snd were safe, and it seemed odd to keep saying "safe"
everywhere.  Now I wish I had made some other choice.  It would be 
possible
to remove all the clutter by exporting a function that sets the safe 
bit;
maybe that's the way to clean this up -- assume unsafe etc.  This issue
came up earlier, and I thought I had found a way to notice in s7 that a
function declared safe was calling something unsafe and warn the user 
(s7_eval,
s7_apply_function, s7_values, and s7_eval_c_string are the culprits), 
but
it turned out to be tricky to trap that case and not also trap innocuous
cases involving calling eval (etc) in scheme.  s7-ffi.html under 
"optimization"
has some discussion of this.  s7.h includes 
s7_define_unsafe_typed_function.

s7_symbol_set_initial_value sets the #_<name> value, so user-defined
functions have that option in hygienic macros and the like.