search-spaces-regexp bunches spaces

[Chong Yidong]

When search-spaces-regexp is non-nil, it does not distinguish between an
individual space and any number of spaces.

To reproduce, create a buffer with the contents

a
a

with no trailing whitespace. Then,

M-<
M-: (let ((search-spaces-regexp "[ \n]")) (re-search-forward "a a"))
[search succeeds]
M-<
M-: (let ((search-spaces-regexp "[ \n]")) (re-search-forward "a     a"))
[search still succeeds!]

This behavior makes it impossible to make a space stand for "exactly one
space or one newline," which may be desirable in text modes. Also, the
values "[ \n]" and "[ \n]+" are treated identically, which is rather
strange.

It would make more sense for *each* space to be replaced by
search-spaces-regexp, if search-spaces-regexp is non-nil.

(search-spaces-regexp is a variable that is new to CVS Emacs.)