NB: Changes to parameter defaults & units included in this commit!

I tried to apply some queuing theory to this problem and as much as
I admire Agner Krarup Erlang, his math isn't much help when you don't
know any of the relevant metrics for your queue.

Instead I took a much simpler approach:  "If we fail to get a thread,
we probably need more threads", and have rewritten the herder to to
react faster and more reliably to such events.

I went over the parameters for thread-pools and normalized timeouts
to seconds rather than milliseconds (beware!) and polished descriptions
etc.

Update test-case to exercise more of the panic formatter.

at a time, paving the way for a lot of future fun...

It's painfully obvious that I set about doing this the wrong way:
I should have renamed sess to req and made a new sess structure,
that would have been much simpler.

Split SES_Handle() and SES_ScheduleReq().

instead of struct sess*

do_stream)

Bug fix itself is not applicable to trunk, but has been pushed on 3.0.

state engine into two separate state engines, one for session
and one for requests.

Next comes the cleanup of all the bits and pieces...

and will make subsequent diffs clearer.

In current automake, there's no way to have the list of tests be
dynamic, so generate that when running configure as well as a local
hook to update the list if it's changed.

Also enable colourised test output.

take a "const struct req *" so enforce this.

Fixes: #1133

into the obscure corners over there, making vmod.std a better copy&
paste example and removing things like "panic" from the VCC.