chore(package.json): Update lib-jitsi-meet.

Broken link, I presume this one was meant.

Updates lib-jitsi-meet to e353322002

Updates lib-jitsi-meet to b8f68b98be
which brings in feature discovery optimization and the unload handler
fix for BOSH.

Updates lib-jitsi-meet to aadfce2283 in
order to fix "not connected" error when jiconop is enabled.

Do so on the main thread at startup and pass it along.

Updates lib-jitsi-meet to 567ba72675 in
order to fix "not connected" error when jiconop is enabled.

Updates lib-jitsi-meet to 8a41d02407 in
order to bring in the XMPP stream resume functionality.

The patch and Prosody 0.11 is required for the XMPP stream resume

https://hg.prosody.im/prosody-modules/raw-file/c49fea05772e/mod_smacks/mod_smacks.lua

The 'previd' query parameter will be use to match user id of the session
being resumed when the smacks module and token authentication are
enabled in Prosody. Otherwise user gets new random id every time and
this doesn't work with the smacks module.

When the reccorder joins, they have a local participant, which is not rendered,
so don't count it towards the partcipant count used for computing the tile
sizes.

Logs the device list when is updated in the reducer and removes
"button enabled" logging which used to dump the device list, but
in a useless way(Object[Object]).

Makes an attempt to log currently selected device, but because of
multiple possible paths it's impossible to find one reliable spot to log
selected device. One has to rely on device list and the GUM call logged
to figure things out.

Due to the outage it wasn't possible to submit them using https://translate.jitsi.org.
Please consider adding them this way.

They just keep crashing.

If the dominant speaker leaves their object will be gone from the mapping.

Regresssion from bd8a7edbd2.

When the toolbar buttons are overridden with URL parameters, our computed set of
buttons will be wrong. Thus, compute it every time and check for the
differences.

It happens when you are drawing into hidden canvas.

On electron 7 drawing on hidden canvas will trigger a memory leak.
Not appending the canvas for the DOM seems to solve the problem.

It's the source of uncountable problems for which we don't have a good
solution, since they are caused by buggy implementations of self-managed
connection services by manufacturers.

To get a workaround in for wired desktop
screensharing in spot in electron 8. With
the change, no "exact" is used in gum
constraints while attempting to get the
wired screensharing device, as that
triggers overconstrainederror.

This effect doesn't modify the media stream, so its safe to start/stop effect and not apply it on the JitsiLocalTrack. This way we can make sure that this effect is not switched out when presenter effect is applied.

Note that in the usual (vertical) setup, the "toolbar" is just the hide button.