Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

It seems to collide, hard, with the (necessary) library mutex.
Also it was really complicating testing.

They tend to produce unwanted effects when an mIRC code is directly
followed by a text digit (which is going to happen often in this context, e.g.
with git hex commit IDs). the mIRC codes have no end delimiter, so
unless the client's recognizer is very carefully crafted it will eat
following text.

- use the right config for Mercurial extractor
- use the default tinyifier if nothing is provided
- support using no tinyifier (=none or empty)

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

I misunderstood the RFCs. Curious that someone else reported this
before I saw it.

Signed-off-by: Eric S. Raymond <esr@thyrsus.com>

This simple change is slightly wrong in a way that IRC servers hide -
strictly speaking we should not append a '\n' to the very last
segment.