For a limited number of cases we can now esi:include a gzip'ed
object into a gzip'ed ESI object.

There is less than 1 chance in hell this will work for you in
practice, I need to handle the N-1 other cases of last bit
and stop bit combinations in the gziped object, but that is
just a small (but tedious) matter of programming.

objects when we know them.

For now: use malloc for gzip instead of running out of workspace.

in an gzip'ed ESI object.

ESI object.

in a gzip'ed ESI object.

included files can get at them.

so we can do ESI, then gzip again to store it.

Add a vef_read() function which implements the esi_syntax&8 bit,
it is too late to do it when we call VEP_parse().

so it applies to all fetch methods.

via sp->wrk->storage.

to tuck in the last storage segment sitting on sp->work->storage,
rather than have this code duplicated througout all the VFP->end()'s

instance into sp->obj.

on reception

ESI+GZIP testcase

and return the offsets we need to compose the VEC string.

The default callback function is a 1:1 mapping.

so we can combine the CRC32 values.

object but wanting to store it ungzip'ed.

processing.

Add separate anchor fields for vgz for reception and esi work.
(The response vgz is a local variable)

Futher modelling of VGZ functions

Move the delivery gunzip away from session workspace and use a
64k stackbuffer for bufferspace instead.