Co-authored-by: Saúl Ibarra Corretgé <saghul@jitsi.org>

Makes the tiles consistent as the participant view is equally sized and the
avatar is always centered in the tile.

Moves the logic from all different places into single state
listener to combine all inputs into a single output.

Up until now we relied on implicit loading of middlewares and reducers, through
having imports in each feature's index.js.

This leads to many complex import cycles which result in (sometimes) hard to fix
bugs in addition to (often) breaking mobile because a web-only feature gets
imported on mobile too, thanks to the implicit loading.

This PR changes that to make the process explicit. Both middlewares and reducers
are imported in a single place, the app entrypoint. They have been divided into
3 categories: any, web and native, which represent each of the platforms
  respectively.

Ideally no feature should have an index.js exporting actions, action types and
components, but that's a larger ordeal, so this is just the first step in
getting there. In order to both set example and avoid large cycles the app
feature has been refactored to not have an idex.js itself.

Use a dimensions detecting root component. The Dimensions module does not
measure the app's view size, but the Window, which may not be the same, for
example on iOS when PiP is used.

Also refactor the aspect ratio wrap component since it can be taken directly
from the store.

Last, remove the use of DimensionsDetector on LargeVideo and TileView since they
occupy the full-screen anyway.

Fixes PiP mode on iOS.

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.

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.

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

* fix(thumbnail): Optimize status bar moderator icon

Moved all moderator functionality to react to optimize the number of
status bar updates.

* fix(RemoteVideoMenuTriggerButton): Use nullish coalescing

Co-Authored-By: Saúl Ibarra Corretgé <saghul@jitsi.org>

* ref(StatusBar): rename to StatusIndicators

* fix(RemoteVideoMenu): isModerator value.

* fix(notification): mobile.

Co-authored-by: Saúl Ibarra Corretgé <s@saghul.net>

Since the verical filmstrip doesn't set its width explicitly anymore,
calculating the available area for the large video based on the
filmstrip width retrieved from the HTML element was wrong
in the cases when the rendering and cleanup of the filmstrip hasn't
finish yet. For example when switching from tile view to stage view.

* ref(Filmstrip): Optimize resizes.

* fix(thumbnails): resize.

* fix(thumbnails): Issue with height 0, width 0.

* doc(Filmstrip): Improve JSDoc.

Simplifies the code a bit, and we use no lifecycle methods.

Audio streams are automatically played by WebRTC and this won't change, probably
ever. There is no point in having checks and an Audio component which does
nothing.

Mobile uses a different logic for deciding whether to show the filmstrip or not:
if there are more than 1 participants or not, and there is no way to manually
toggle it.

This refactors all handling of audio-only and last N to 2 features in preparation
for "low bandwidth mode".

The main motivation to do this is that lastN is a "global" setting so it helps
to have all processing for it in a single place.

The filmstrip is currently only hidden when PiP mode is entered. There is no
real reason not to leave the large view as it was.

🔥🔥🔥

Using anything non-serializable for action types is discouraged:
https://redux.js.org/faq/actions#actions

In fact, this is the Flow definition for dispatching actions:

declare export type DispatchAPI<A> = (action: A) => A;
declare export type Dispatch<A: { type: $Subtype<string> }> = DispatchAPI<A>;

Note how the `type` field is defined as a subtype of string, which Symbol isn’t.