[CM] R7RS in S7

[bil at ccrma.Stanford.EDU]

Good point; I think s7 could have a 'debug field
or something, and if set, could use big cells (already
in use for other special cases); currently cells
take 48 bytes (a miracle of bit-fiddling compression,
if I may say so), but gensyms take 56, so the facility
exists to tack on other fields.  The line numbers
are sometimes lost because that portion of a cell
is also used by the optimizer; expansions are
a special problem -- they are read-time operations,
so the current line-number (file-wise) isn't changed.
And it's really unclear to me how much info to
put in the circular history buffer -- too much and
s7 slows down for no reason, etc.  I did recently
fix a lot of the line-number problems -- it turned
out to be a dumb typo in s7_error.