Implement ellipsis escaping, as per R7RS

[jpellegrini]

In Section 4.3.2 of R7RS (page 23 of the PDF), it is said that a can be one of:

1. (<element> ...)
2. (<element> <element> ... . <template>)
3. (<ellipsis> <template>)
4. #(<element> ...)

Number (3) is an escaping pattern. Inside ir, ellipses are treated like a symbol, and not like the syntax-rules ellipsis. And that template must have only two elements.

(<ellipsis> <template>)
should be transformed into
<template>

but protecting the ellipses (they have no special meaning there).

So

(... a)          => a
(... (a b ... c) => (a b ... c)
(... ...)        => ...
(... a b)        => error (too many elements)

but the ellipses on the right side are ignored by syntax-rules.

The implementation is:

1. In MBE, include a check for malformed escapes (as in the last case above, (... a b))
2. In MBE: (mbe:subst-ellipsis expr ell new-ell) will go through expr, and substitute only the sublists that match the form of an escaping ellipsis, doing the appropriate transform. However, inside the subtemplate, the ellipses will be transformed into a symbol which is internal to MBE, produced by gensym (so no match will be possible). This is in find-clause
3. in MBE, "replace" can be used to get the original ellipses back. This is also in find-clause
4. In runtime-macros, syntax-rules calls the internal MBE procedure check-wrong-ellipsis-escape
5. In MBE, let-syntax also does that.

Some tests, including the example in R7RS, were added.

Fix #704

[jpellegrini]

There is a glitch in this, I'm fixing.

[jpellegrini]

Ok, I got it working. But one of the examples does not work with let-syntax. I think it's something in let-syntax; I'll take a look later. Its code seems different from that of syntax-rules