PATCHES: richer listtext

Hi, all,

Given there has been positive interest, and there will be minimal/no
disruption to mlmmj's operation, I have implemented the richer listtext
features I wrote about earlier.

I'm sending these patches through now, though they have had only minimal
testing, particularly as Robin Johnson has volunteered to help test.

mlmmj-listtext-headers.patch
- Allows arbitrary headers to be included at the beginning of listtexts.
- All the headers in the listtext will be included in the mails mlmmj
   sends.
- Mlmmj standard headers of the same name as those in the listtext will
   be omitted.
- Substitutions using the $whatever$ mechanism can be used.
- \uNNNN escaping is supported in the headers, as with the rest of the
   listtext.
- Automatic =?utf-8?q?...?= quoting is done for (and only for) the
   Subject: header.
- Also includes a couple of bugfixes related to sending digests.
- This is compatible with current behaviour.
- The interface to prepstdreply() is changed, however; there is no
   longer a customheaders argument. It was never used, and since this
   patch allows custom headers to be included in listtext, is not really
   necessary as a function argument. Incorporating it in the
   implementation would have been more trouble than value.

mlmmj-origmail-count.patch
- Alters the way $originalmail$ works; it doesn't buffer the entire
   $originalmail$ substitution in memory, but does it a line at a time.
- $originalmail$ must be first on a line, optionally preceded by
   whitespace. The whitespace is prepended to each line of the mail that
   is included.
- The rest of the line following $originalmail$ is ignored.
- $originalmailNNN$ can be used, where NNN is a number of lines to
   include. Use a large number such as 1000000000 to include the whole
   mail (but be aware of integer overflow so don't go too big!).
- If NNN is omitted, the default is 100.
- To get current behaviour, a space must be prepended to the lines
   currently containing $originalmail$ in the listtexts. Apart from that,
   this is compatible with current behaviour.
- Again, there is an interface change. substitute() and substitute_one()
   no longer take the original mail filename, and cannot be used to do
   $originalmail$ substitution. This was never used except by
   prepstdreply() which now incorporates that substitution itself.

mlmmj-subject-substitute.patch
- Allows $subject$ to be used for the subject of the mail being
   moderated (or denied, etc.).
- Compatible.

mlmmj-random-substitute.patch
- Allows $random0$ through $random5$ to be used as distinct random
   strings.
- Compatible.

mlmmj-efficient-random.patch
- Makes random number generation more efficient by only seeding the
   generator once.
- Compatible, though behaviour will change slightly.

mlmmj-fixed-length-random.patch
- Makes the random strings produced always the same length rather the
   smaller random numbers producing shorter strings which could be
   problematic.
- Compatible, though behaviour will change slightly.

I still plan to include proper documentation (in the patches or as a
separate patch), but for testing purposes, this email should suffice.

The moderation listtext I have tested with is attached.

I believe the patches will not all apply cleanly unless you have already
applied these patches:

- patch-mlmmj-1.2.17-origmail.diff (mlmmj+get-1844_at_mlmmj.org)
- mlmmj-notifymod.patch (mlmmj+get-1826_at_mlmmj.org)

(I've attached them for convenience, since one of them doesn't seem to
have made it into the web archive, and the +get trick hasn't worked for
me yet, though I suspect the latter problem is just a delay.)

Nor will all apply cleanly unless you apply them in order.

If you have trouble, let me know. It's possible another of my patches
I have applied affects this, though I don't think so.

The following is a rough list of things that should be tested that I
made as I implemented:

- digests with and without thread summary work
- Content-Transfer-Encoding headers correctly output for both digest and
   listtext
- multiple headers of same type in listtext are preserved
- \uNNNN escaping works
- number of lines of $originalmail$ correctly honoured
    - when larger than mail length
    - when smaller than mail length
    - when equal to mail length
- number of lines of $originalmail$ handled as 100 when not given
- whitespace preceding $originalmail$ correctly prepended to every line
   of included mail
- content after $originalmail$ ignored (no crash)
- six random values are distinct, but the same when reused in the same
   listtext
- subject substitution
    - including when original subject contains utf8
- headerless listtext (warnings output)
    - with and without blank line
- invalid header in listtext OK (no crash)

Help testing will be particularly appreciated. Code review and other
feedback also welcome.

Ben.