Only available in vcl_backend_fetch{}. If it is unset varnish will not
pass the request body to the backend.

This is in preparation for addressing #1927.

When handling a fetch error we may not even have had memory to store
the status code.

Fixes #1990

Relates to #1990

Fixes #1994.

soon need them.

Fixes #1989.

actually possible to locate the compiled VCL shared library based
on the loaded VCL name, and removed the part of the subdirectory
name which made it unique.

Bad idea.

Add a comment to make sure I don't get that Idea again.

Fixes #1933

Spotted by: Flexelint - phk yeah \o/

- tables get included multiple times
- we use do ... while(1) all over the place
- dont always include config.h for flint, source files need to include it
  so we want to see warnings if we missed to include it
- tell flexelint where config.h is ok not to be used (sources with
  no actual code unless a feature is HAVE_ defined)

Test by scn@. OK by @phk.

Fixes #1963.

The counter g_cooling variable is now decremented when memory from the
cooling list is freed, making the counter vsm_cooling update
accordingly.

Fixes: #1984

This new parameter specifies how long freed VSM memory is on the
cooling list before it is actually freed.

Waiters hang off the objhead, so the simple fact that we manipulate
some objcore on the same objhead does not imply that there are
relevant news to the waiters.

We now are more specific about when we wake up waiters and how many.

Never for
- expiry / ban / purge / nuke
- stale objects we just happen to reference

We do want to rush when
- fetching a busy object failed or got aborted: We should wake up one
  waiter to re-attempt the fetch
- we succeeded fetching a busy object or delivering: Rush more waiters
  by policy

This fixes the general issue described in #1928 that we should not
rush when expiring.

Closes #1945