Switch media processing to conference bridge architecture

[MayamaTakeshi]

Better do this before #22, #56 and #57.

[MayamaTakeshi]

And before doing this, we should solve #66 so that we can prepare a test with something like 256 simultaneous calls checking DTMF in all channel so that it can be used to confirm there was no negative impact in performance after the change.

[MayamaTakeshi]

We should study the code in https://github.com/pjsip/pjproject/blob/master/pjsip/src/pjsua-lib/pjsua_aud.c It indicates we create a single pjmedia_conf to handle media for all calls.

And this is confirmed by nanang:

Just an opinion, I don't see any advantage of creating conf bridge for each call, since it's just wasting resources (and perhaps increasing code complexity). So I would suggest you an alternative design:

• For all calls: a shared conference bridge driven by master port/null audio.
• For each call: just create a separate memplayer port, add conf port for memplayer port & portChannel port to the shared conference bridge, and connect them.

FYI, pjsua-lib also uses a single conference bridge to handle multiple calls simultaneously. (Btw, why not just using pjsua-lib for your app, it should be simpler).

https://pjsip.pjsip.narkive.com/93IWjFGF/handle-mutiple-media-flows-simultaneously

Also, check: https://github.com/pjsip/pjproject/blob/5ef8463d562db34adaf5ba6969b50d3814f3939f/pjsip/src/pjsua-lib/pjsua_call.c#L5273 and https://github.com/pjsip/pjproject/blob/master/pjsip/src/pjsua-lib/pjsua_media.c#L3764 https://github.com/pjsip/pjproject/blob/master/pjsip-apps/src/samples/confsample.c

and maybe: https://github.com/Ardameon/experiments/tree/master/pjsip_inv_example

[MayamaTakeshi]

I tried using a single pjmedia_conf but samples/100_calls.js got stuck when adding tonegen ports to it. So I tried with one pjmedia_conf per call and after that the issue was solved.

[MayamaTakeshi]

Test evidence:

takeshi@takeshi-desktop:sip-lab$ time ./runtests 
... ABRIDGED ...
Success. All tests passed

Successful tests:
  - samples/100_calls.js: duration=57.30 seconds
  - samples/16_audio_streams.js: duration=4.68 seconds
  - samples/183_session_progress.js: duration=7.91 seconds
  - samples/delayed_media.js: duration=5.20 seconds
  - samples/four_audio_streams_two_refused.js: duration=3.70 seconds
  - samples/g729.js: duration=0.99 seconds
  - samples/mrcp_and_audio.js: duration=1.14 seconds
  - samples/mrcp_and_audio.simplified_media.js: duration=2.39 seconds
  - samples/options.js: duration=0.18 seconds
  - samples/refer.js: duration=0.87 seconds
  - samples/refuse_telephone_event.js: duration=1.90 seconds
  - samples/register_no_expires.js: duration=0.18 seconds
  - samples/register_subscribe.js: duration=0.59 seconds
  - samples/reinvite_and_dtmf.js: duration=4.61 seconds
  - samples/reinvite_audio_audio.js: duration=4.90 seconds
  - samples/reinvite_with_hold_unhold.js: duration=8.46 seconds
  - samples/rtp_and_srtp.js: duration=5.60 seconds
  - samples/rtp_and_srtp.rtp_refused.js: duration=5.30 seconds
  - samples/send_and_receive_fax.js: duration=63.97 seconds
  - samples/simple.js: duration=0.47 seconds
  - samples/sip_cancel.js: duration=0.44 seconds
  - samples/srtp.js: duration=5.30 seconds
  - samples/tcp_and_extra_headers.js: duration=9.94 seconds
  - samples/two_audio_streams.js: duration=5.72 seconds
  - samples/two_audio_streams.port_zero.js: duration=5.71 seconds

Everything OK

real    3m27.539s
user    0m20.691s
sys     0m8.487s
takeshi@takeshi-desktop:sip-lab$