VMODs should never, ever, see.

can also use it.

Patch by:	martin
Some rework by:	phk

Patch by:	martin	(slightly polished)

Redhat packaging files for the official varnish-cache.org
RPMs will from now on live in a separate git repository.

This decouples source releases (including release tagging) and
the packaging work, which makes the release process easier.

Instead of asserting, return a 500 error to the client if
running out of client workspace.

This also extends vmod_debug with workspace handling functions, and
adds two test cases that use these functions.

As discussed at VDD15Q3. Reviewed by Martin.

Fixes:	#1782

I often have to look up which one is which, try to avoid
that in the future.

The Varnish CLI documentation was advertising an optional temperature
parameter for `vcl.use`, and overall the test coverage wasn't deep
enough. There is still room for improvement.

The debug VMOD now logs some of the VCL events to help cover more
behavior in `v00044.vtc`.

This reverts commit d771942c, it needs
to be discussed first.

OSX/64 issue...

Related to: #1773

we don't yet know the length.

I attempted to add a way to determine if the object was big enough
*yet* to satisfy the Range, but that was non-viable.

Fixes:	#1777

Don't install cache_backend.h and cache_director.h, they expose more
than what a VMOD needs to see.

The VRT API jumps directly from struct vrt_backend to struct director,
which makes struct backend more than simply opaque, it is completely
invisible. However struct director needs to be visible and has been
moved to cache.h along with its functions definitions.

... like 'Unknown keyword: $Init' for 4.0 vcc files.

Make sure probes can be managed outside of the cli thread, now that
backends can be created any time.

Fixes:	#1775