This drastically reduces the number of printf-like operations.

Fixes #3968

And use it to implement ALLOC_FLEX_OBJ().

Even though the flexible array may start inside the padding, an array of
zero elements could then lead to less than sizeof bytes.

We lose the ability to check that fld is indeed the last field, but we
never had that ability to begin with.

We did mention it in the release docs, but not in changes.rst

We must ensure that we get a complete header name before evaluating
a ban, so "req.http." should not be accepted.

Refs #3962

Ban expression variables that partially match standard variable names
(ex: req.urlXX) should not be accepted, except for variables that take
an HTTP header name as a suffix.

Fixes #3962

by default, add "User-Agent: cNAME" header to all client txreq requests
if not already present (via -hdr). This can be disabled with -nouseragent

by default, add a "Server: sNAME" header to all server txresp responses
if not already present (via -hdr). adding -noserver to txresp will
disable it. Note that for -dispatch servers, the id of the spawned server
(sN) will be used instead of s0.

Fixes #3960

Suppose the following happens:

vbp_task() finishes with vt->running = 0 and a heap insert. The
vbp_cond is signaled under the lock, but now instead of vbp_thread()
waking up first, VBP_Remove() gets the lock and reaches
assert(vt->heap_idx == VBH_NOIDX) before the racing vbp_thread()
deleted the heap.

This is unlikely to happen with static backends, because for those,
the probe is stopped via the vcl temperature before they get removed.

Fixes https://github.com/nigoroll/libvmod-dynamic/issues/100

And mark it for flexelint

my local config on linux is slightly different than phks on freebsd