Re-use plotly plot if config did not change

[falkoschindler]

This PR solves #1386 using Plotly.react instead of Plotly.newPlot when the config did not change. This allows using uirevision to keep the user state (zoom level etc.) while updating the data.

The distinction whether the config changed is necessary because Plotly.react behaves badly if we pass options.config with responsive=true being set. But since it doesn't change when only updating data, we can still use the more efficient Plotly.react in these cases.

While being at this, I improved the update mechanism (implement the updated method instead of requiring to call update from the server) and moved the default setting of responsive=true to the client (simplifies the Python code, combines all config trickery in one place).

This test code provides a button to update the data as well as one to update the config:

from random import random

fig = {
    'data': [{'x': [1, 2, 3], 'y': [0.5, 0.5, 0.5]}],
    'layout': {'uirevision': True},
}
with ui.row().classes('bg-red-100 p-2 w-full'):
    plot = ui.plotly(fig).classes('w-full')

def add_trace():
    fig['data'][0]['y'] = [random(), random(), random()]
    plot.update()

def show_display_mode_bar():
    fig['config'] = {'displayModeBar': True}
    plot.update()

ui.button('Update', on_click=add_trace)
ui.button('Show display mode bar', on_click=show_display_mode_bar)

The plot remains responsive (i.e. fills the width of the container when resizing the window) and keeps the user state (e.g. zoom level) when updating the data. Only when the config changes (e.g. displayModeBar) the user state resets because the plot is re-created.