new packet format: dictionary instead of positional arguments

[totaam]

We can support both the old format and the new one for a while.
The new one will have the advantage of being easier to understand and more flexible so we can add new features, make some parts optional, without having to fill in positional arguments we may not have. (they may not be relevant or may not be useful)

[totaam]

Example of ugly packet formats which would benefit from this: r20252 (adding relative position to the existing position parameter to avoid adding it at the end of the packet, which has optional fields..), configure packets, map packets, etc...

We can continue to support the old format, and convert to the new one - and maybe the same conversion function could be used by both the client and server to keep the code simple.

[totaam]

Could also be used for keyboard (#1172, #1716) and refresh (#1449), pointer events (#1941)
On a more general note: info packets and #1983, moving keyboard packets out of hello, hello namespace (#1154) are all related to legacy formats.

Could be useful for #2243#comment:1

[totaam]

Started for pointer-position so that we can also enable QUIC (#3376) datagrams for those packets easily: just use a new, cleaner packet handler for it.

Still need to:

• populate server_packet_handlers with the (full?) list
• use a similar mechanism for the server?
• continue converting the other pointer and wheel events and anything that calls process_mouse_common, ie: window-configure
• is there any point in sending button events as both datagrams and regular packets? (I think not - perhaps during heavy congestion?)

[totaam]

5178602 caused problems so the new packet types are now disabled until I find the time to fix: 284aeec

[totaam]

New:

• wheel events should be sent raw, not as pairs of clicks
• appindicator has a scroll event

[totaam]

Ideally, we can use type hint introspection to move the packet data validation to a generic function that enforces the data types early.

Helps with #3618, see also #3981