Seen on a solaris vintage edition (snv_111b) as SIGSEGV caused
by strlen(NULL) in r01036.vtc and r01037.vtc: a z_stream msg
can be NULL.

Changes:

* 302 Found
* 307 Temporary Redirect

  Do not apply the default ttl, only set a ttl if Cache-Control or Expires are
present.
  Responses with these status codes are not cacheable by default

* 414 Request-URI Too Large

  Cacheable with "heuristic expiration"

https://www.varnish-cache.org/patchwork/patch/251/

This hopefully makes l00002.vtc less prone to error.

over the weekend:

Just because we have a bo doesn't mean we have a bo->beresp any more,
find the C-L header in the req->resp.

64bit processors.

We have to restart the server instance when we have writes which
might fail.

are handled so fast that there may be no numerical difference between
the start and end session timestamps.

This is a consequence of our decision to use C type double (ieee 64bit FP)
for timestamps, at the current distance from POSIX epoch, 1.5 billion
seconds, the resolution is 238 nanoseconds.

Patch from daghf (with minor changes)

Submitted by:	fgs

as priv's inside ESI's delivery code.

delivered.

Transports are a superset of protocols because internal requests,
notably ESI:includes, doesn't use a HTTP protocol to deliver their
response, but have have ødipal ways of doing things.

Submitted by: Dridi

it to slim down req/busyobj a tiny bit.

its cheaper to eliminate the exceedingly unlikely resource leak
Coverity spotted that way.

Fixes #1692

Found and diagnosed by:	martin
Slightly different patch by me.