When STV_BanInfo fails, it means the stevedore has run out of space to
persist the list. An export should descrease the size.

this on startup, after reading the persisted ban lists, to make sure
all stevedores has all the bans from every stevedore stored.

until ban compile.

Document the reason.

reload all the statements in one go.

makes sure that the ban lists will not change until we exit.

This is to give persistent stevedores a chance to clean up their ban
lists without having to worry about the lists changing under them.

The subsequent cleanup/separation of .h files makes this test
much less important, and it's not like it is a conclusive or
comprehensive test in the first place.

Fixes	#1230

Output that we've switched VCLs and which one we have switched to when
using vcl.use.

Fixes #1232

Fixes #1233

This commit included som unfinished work, so revert it.

This reverts commit d4f456f0.

Fixes #1233

ban.url is confusing as it takes a regular expression rather than a
fixed string, so get rid of it in favour of people being explicit and
using ban req.url ~ /foo

10 loops.  Simplify the structure of the code some more.

Prodded by:	martin

rst2man doesn't support the `glossary` markup, it's a sphinxism.
Avoid using that and sort the list manually instead.

Merge virtualised into a new platform specific notes section. Add transparent hugepages info from ticket 1054.

Original patch by:	martin

don't forget about it so it hangs around with the ban, doing nothing.

(This can happen with /etc/hosts)

Inspired by patch from:  Tollef

Prodded by:	Martin

backend VXIDs to request VXIDs.

Still TBD: where ESI fits into this.

belong to the same batch of VSL records and therefore have the
same vxid.

This will allow us to optimize libvarnishapi, in particular when
doing ordered/transaction mode.

This resolves the race where the pools might accept connections and
start processing them before a lot of the initialization routines has
finished.