Make sure subrequests to not see Content-Length / Transfer-Encoding

In other news: I've gotten much better at writing python.

Otherwise forks may attempt to run coveriy scan builds only to fail and
spam their owners.

Only momentarily until we understand and fix the newly introduced
issues.

This reverts commit f0ee94ec.

Ref #3813
Reopen #3761

There is no cancellation mechanism in the middle of an HTTP/1
transaction besides closing the session. If a backend closes
the connection before the end of the bereq delivery, it may
also send a 4XX or 5XX response that could be valuable for the
original client.

There are several things to take into consideration: it is
the responsibility of the backend implementation to raise the
bo->send_failed flag, if the backend replied something before
closing we shouldn't trigger a timeout, the connection must not
be recycled, and the object must not be cached.

Fixes #2332
Closes #3686

With the help of PARAM_PRE and PARAM_POST macros we can deal with the
include tables bits included by the params include tables without the
nested macro evaluation limitation.

This will allow the conditional definition of parameters in multiple
steps, which is needed by bits parameters.

Signed-off-by: Dridi Boukelmoune <dridi.boukelmoune@gmail.com>

For other VFPs, we use the scheme

	<vfpname>_<callbackname>

It differs for vtest running on IPv6

Bring back the positive test, testing with a chunked body one
byte less, exactly and one byte more (than) the cache size.

In d7695e48, a legit caching case was turned
into an error case.

Allowing units is much more convenient and keeping them optional
maintains compatibility. On the other hand, changing the defaults
for default_ttl and default_grace has no impact on the generated
RST.

Keeping track of the user input could help since default values
are "washed" like user input, without resorting to a dynamic
default.

Refs #3756

By themselves, structured fields parsing functions don't fail on extra
input. To avoid repeating the same checks twice, they are wrapped into a
function that will help generalize tweak_generic_double() at least once
more.

This is a first step towards even more generalization.

TCP_KEEPALIVE is the TCP_KEEPIDLE's equivalent and uses seconds
 for the idle time.

This session close reason is also used for receiving response
bodies, for example:

--  TTL            RFC 86400 10 0 1652368707 1652368707 1652368707 0 86400 cacheable
--  VCL_call       BACKEND_RESPONSE
--  VCL_return     deliver
--  Timestamp      Process: 1652368707.157422 0.063369 0.000054
--  Filters
--  Storage        file s0
--  Fetch_Body     2 chunked stream
--  FetchError     chunked read err
--  BackendClose   53 self close Failure receiving req.body
--  BereqAcct      1337 0 1337 987 3465216 3466203
--  End

so #ifdef __FreeBSD__, which is OK, since the intent is just to get
VTEST code coverage.

Closes #3795

if only for consistency.

I will keep an eye on vtest.

This reverts commit cea729bb.