Improve event registration

[falkoschindler]

This PR attempts to improve the generic event registration with the .on() method.

• provide all event arguments by default, e.g. clientX for "click" on a label
• let .on pass information about sender: add msg: Dict[str, Any] to EventArguments
  .on Event for Element doesn't pass information to callback about sender #583
• improve access to event arguments, e.g. row and index for "rowClick" on a table
  Improve access to event arguments #672
• allow filtering more arguments

[falkoschindler]

I found a way to define, which event arguments should be included, that is easy to use and backward-compatible:

1. A list of strings names the attributes of the JavaScript event object:

   ui.button().on('click', handle_click, ['clientX', 'clientY'])

2. An empty list requests no attributes:

   ui.button().on('click', handle_click, [])

3. The value None represents all attributes (the default):

   ui.button().on('click', handle_click, None)

4. If the event is called with multiple arguments like QTable's "row-click" (evt, row, index) => void, you can define a list of argument definitions:

   ui.table(...).on('rowClick', handle_row_click, [[], ['name'], None])

   In this example the "row-click" event will omit all arguments of the first evt argument, send only the "name" attribute of the row argument and send the full index.

If the retrieved list of event arguments has length 1, the argument is automatically unpacked. So you can write

ui.button().on('click', lambda e: print(e.args['clientX'], flush=True))

instead of

ui.button().on('click', lambda e: print(e.args[0]['clientX'], flush=True))

Note that by default all JSON-serializable attributes of all arguments are sent. This is to simplify registering for new events and discovering their attributes. If bandwidth is an issue, the arguments should be limited to what is actually needed on the server.

[rodja]

Looks super cool. I think we should add some in-detail documentation about the .on event registration system including your explanations from #1095 (comment).