http_resp_size is allocated on the thread_pool_workspace.  This means
we will hit assert errors if http_resp_size is larger than the
thread_pool_workspace.  8k should be enough in most cases, from a
light sample of web sites in the wild.

Thanks to askalski.

closes #867

Keep is for future use by conditional backend requests.

req.ttl provides a way to force a fetch of a not yet stale object,
see how in testcase c00042.

We can check the size of the return value too.

I forgot to rename the tv_get variable to tv in configure.ac

Fixes #853

(again)

Fixes: #853

it no longer saves us memory.

rather than saving the entire thing.

This also moves a bit more of the fetch-policy/standards-compliance into
rfc2616.c

accordingly.

initialize struct exp, and use it.

Pass an 'exp' instead of just the ttl to STV_NewObject()

Have -spersistent always use the graced timeout for object index,
rearm when we resurrect the zombie object.

in cache_expire.c

All this interior decoration should make it less painful to add
yet a timer for IMS processing.

manipulation is actually the relative value, not the absolute time.

HSH_Grace() to EXP_Grace()

Found & Patch by:	Dmitry Panov

in order to encapsulate the lack of of API coordination between
different OS's in this arena.