Before the h2 frame dispatch we only need to check that coming from a
HEADERS frame the very next frame is a CONTINUATION, when the HPACK
block didn't fit in the former.

The CONTINUATION dispatch will then make the stream consistency check.

When a stream times out waiting for window credits, and all the other
streams are broke, declare the whole connection bankrupt.

It replaces idle_send_timeout when the stream is waiting for window
credits before sending more DATA frames. The idle_send_timeout will
still apply to individual writes to the socket, but triggering it is
considered a failure condition (this was already the case).

The two loops are merged into a single one to better deal with the
lack of ordering guarantees of request vs connection stream control
flow window crediting.

Refs #2980

This parameter needs to be somewhat high because web browsers may for
example only credit streams for image resources just enough to parse
metadata like width and height of a picture in order to perform layout
computations and fetch resources more sensitive to latency before the
effective image payloads.

Conflicts:
	include/tbl/h2_error.h

The new meaning for h2->sess->t_idle is now the time since the last
complete frame, except before the first frame where it corresponds to
the creation of the session.

And after issuing or receiving a goaway, stop processing frames when
there are no streams left.

Initially submitted as part of listen sockets management.

Refs #3959

Since Python 3.12 unknown escape sequences trigger a warning.

The 503 synth and 500 minimal response status codes are too misleading
in this context, where the failure is attributed to the client. Among
existing 4XX status codes, this is the closest if we stretch the timeout
definition to "didn't complete rapidly enough before the client went
away".

A VCL failure on the client side transitions to vcl_synth, except
failures from vcl_synth that lead to minimal errors. The ESI transport
is not allowed to reply with minimal responses so this would lead to a
panic.

On top of that, the vcl_req_reset feature flag emulates `return (fail)`
statements when an HTTP/2 client disconnected, resulting in the same
panic scenario.

For sub-requests, we masquerade the fail transition as a deliver and
trade the illegal minimal response for the synthetic response.

Fixes #4022

Conflicts:
	doc/changes.rst

Conflicts:
	doc/changes.rst

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

Ref: 582ded6a

Fixes #3958

The VSB bits in here come with a very significant performance penalty
for H2 request body processing.

This requirement was dropped in the updated rfc 9113.

Fixes: #3911

with uds abstract sockets, sun_path should start with a NULL character followed
by the socket's name. The name is not considered to be NULL terminated and can
contain NULL bytes which have no special meaning. socklen is used to determine
the length of name and must be set to the length of the struct sockaddr_un up to
the last character of name, otherwise the 108 characters of sun_path will be
treated as the name of the socket, including NULL bytes.

Adds coverage for tab characters at start/end of field value.

Regarding the "fo o" " bar" header, it cumulates an error in the name
and another in the value, but only one of them will trigger the expected
PROTOCOL_ERROR. Only the invalid "fo o" is checked now, and the other
error is part of the new coverage.
Signed-off-by: Dridi Boukelmoune <dridi.boukelmoune@gmail.com>