otherwise that's pretty useless in SI flow

fixes regressions by commit
7e9f4ed3
(early multi sections table end, segmented EIT failure)

Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

The function dvbpsi_chain_demux_delete() failed to call p->pf_del() for all
attached decoder. This resulted in leaked decoder memory.

SCTE35 / CUEI Descriptor parsing bug. Descriptor lengths of one byte fail to parse, instead returning null.

The Spec from 2004 to 2016 shows the descriptor payload has never been
more than a single byte.
See SCTE 35 2016 Page 19 Table 2 (Section 8.2).

Correct the length validation check, and reject any descriptor that
IS NOT one byte long, as per the spec.
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

src/demux: functions dvbpsi_*DemuxSubDecoder() are no longer implemented, but are merely empty stubs

The function dvbpsi_decoder_psi_section_add() never set p_decoder::i_last_section_number when
adding a new section at the end of the list. This was left to the decoder implementations. However
the correct action would have been to set p_decoder::i_last_section_number in dvbpsi_decoder_psi_section_add().

The EIT tables were output incomplete as the
first sections were always dropped.

(2 sections, last_section == 1)
Ex: section 1
    section 0
    secton  1

IsCompleteEIT is always true in first condition in that case.
It is also always true if section is N/N.
first_section > -1 && section_number == last_section because
it has just been added.

Doing the check before effectively adding the section
allows a correct evaluation against current section.
Signed-off-by: Jean-Paul Saman <jpsaman@videolan.org>

Using the current date for capture time makes no sense when the source is a file.