This commit removes dependency on ChatRoom and JingleSessionPC from
the TraceablePeerConnection.

* re-enable setting for enabling/disabling rtx

* add tests for rtxmodifier.  (and a bugfix found by the tests)

* remove rtx handling from sdp-consistency (it will be handled separately)

* better error printing in JingleSessionPC around renegotiation

* add rtx modifier to handle inserting local rtx ssrcs, keeping local rtx ssrcs consistent, and imploding remote rtx ssrcs

* jsdoc, es6 cleanup, logging tweaks

* fix line that's too long

* add a 'parseGroupSsrcs' helper function

* fix a bug when passing group ssrcs to be parsed

* address pr feedback

* address pr feedback

* fix for loop

* address pr feedback

* better const in SampleSdpStrings

Video SSRC has to be generated for the recvonly stream if there are no
video tracks in the PeerConnection.

Fixes crashed when null stream (muted) is being removed.

Add 'replaceTrack' method to API, make all replace track operations local (use same ssrcs so no need for source-add/source-remove)

* M1: device change now uses new flow.  fundamentally "works" but may be corner cases/side effects to other flows. haven't touched ffox yet

* M1a: fixed a bug where the ssrc info of the new stream wasn't being set correctly, causing problems with video mute/unmute

* M1b: fix firefox.  changing camera, video mute/unmute looks good

* M1c: 'replaceStream' processing is not integrated with the same queue as all of the other o/a-related events.

* fix typo, some deprecated code cleanup

* move the rfc4145 fix to traceablepeerconnection::createanswer so both the new and old o/a flows leverage it.  also cleanup some now-unused code and change more functions to arrow functions so we can get rid of 'var self = this' workarounds.

* get rid of separately-mainted local description in JingleSessionPC.  keep TraceablePeerConnection as the source of truth for the local sdp and have JingleSessionPC retrieve it from there.

* change the code in the old modifySources shim to use the _renegotiate method so both old and new flows use that method.

* split out the remote change processing from the local change processing to better set us up for individual remote and local modification queues

* move the setOfferCycle flow off of the old shim and into the new flow.  also removes the new for this this.jingleOfferIq member

* move addRemoteStream (nee addSource) to the new flow.  gets rid of the need for the this.addssrc member.

* move remove remote stream over to the new flow.  gets rid of the need for the this.removessrc member variable

* temporary '_processRemoteChange' method is no longer needed, as all remote changes are processed directly (rather than setting member variables and calling a catch-all method which checks each member to see what needs to be done)

* M2: add 'sdp-consistency' in an attempt to separate out the restoration of the same ssrcs (also requires the changes i've made to sdp-simulcast, currently in my brian/rework branch @90aceaddb1eec4d15026efac82c140579c7af2b3, as well as a commit in jitsi-meet to change the toggle-screenshare flows over to use the new 'replace' chain).  as of this point video mute, change device and toggle screenshare all happen as local operatons.  there is a bug when starting as video muted i need to look into, and i haven't checked firefox yet.

* fix the bug with start-as-video-muted (relies on commit f1d8d3c831 in brian/overual sdp-simulcast branch)

* some code cleanup, marked some TODO list items.  got addStream off of the old shim and onto the new flow

* change removeStream over to the new flow and off of the shim.  shim is now removed!

* fix a bug when muting a stream.  remove some logic from addstream/removestream that i don't think makes sense, and move the source-remove parsing logic into a helper

* move the source-add/source-remove processing into the atomic queued operation since it looks at sdp.

* remove unused code used for old ssrc replacement logic

* get rid of the need for TraceablePeerConnection::addStreamNoSideEffects

* get rid of need for TraceablePeerConnection::removeStreamNoSideEffects

* distinguish XXXStream and XXXStreamToPeerConnection (was XXXStreamNoSideEffects)

* remove shim methods for add/removeRemoteStream

* fix a bug when adding local media for the first time

* replace the old 'dispose' method with the new one

* address feedback comments.  update version of sdp-simulcast.

* update sdp-simulcast version

* add missing semicolons

* move sdpconsistency to es6 style

* restore JitsiConference::removeTrack, restore JitsiLocalTrack::dispose back to how it was (doing o/a if necessary)--we get around this on replace by not disposing the track until after we do the replace)

* change removeTrack and addTrack to leverage replaceStream

* fix removeTrack to return the promise, handle an already-disposed track

* add more documentation

* remove unnecessary semicolon

* update sdp-simulcast version

* address PR feedback

* remove usage of @type in method js docs

* address (more) pr feedback.  upgrades to a new sdp-simulcast version which fixes an issue in smoke tests

* more PR feedback, and clean up the success/error handling done for the work functions.

* tweak the error propagation

Currently if RTCPeerConnection is closed (an event is delivered) and we haven't called close method ourselves, this means browser has closed it due to suspending the PC.

ICE connectivity issues can be transient and those should not prevent
from sending SSRC updates. XMPP stream is a separate connection/stream
and skipping the updates while ICE is disconnected results in broken
media session where the stream information is not propagated correctly
across the conference participants.

Introduce JingleSessionState constants. Make the invalid JingleSession
rejection an explicit action taken by the JitsiConference.

The flag when set to 'true' will enforce ICE failure. It is supposed to
be used in automatic testing.

Our JSHint configuration is not extensive and we have excluded multiple
files from linting. Additionally, we have seen JSHint to be unable to
parse newer ECMAScript features such as the object spread operator
(proposal).

On the other hand, we have seen ESLint to beautifully work on React and
React Native source code in addition to ES2015 and later. Introduce
ESLint alongside JSHint as an intermediate step to eventually switching
from JSHint to ESLint.

As our source code does not fully follow even the JSHint rules we have,
it is very difficult to introduce ESLint with many rules. At the
beginning, introduce it with as little rules as possible. Morever,
ESLint is able to automatically fix certain rule offenders so once we
have ESLint merged we can incrementally add new rules while keeping our
commits' modification on as little subjects as possible.

Do not flush the connection after 'session-accept'
is queued to avoid BOSH race condition issue.

This is an unrecoverable error and in current state of
the system reload in the client is the least complex
thing to do.

Moves render time calculations in the library.

The previous method was clobbering the JingleSession prototype,
preventing the creation of additional subclasses.

When source-add and source-remove for the same SSRC arrive just one after
another, SSRC will not be removed as it didn't make it to the remote description
yet.