This PR passes observed=True to omit categorical values with no observations in the dataset. Fixes #4274 and #4433.

Also silences the FutureWarning from pandas-dev/pandas#43999 since pandas 2.1 that the default of observed=False will be changed to True in a future version.

Following code raises KeyError: 'Fri' before this fix:

import plotly.express as px
df = px.data.tips()
df["day"] = df["day"].astype("category")
px.scatter(
 df[df.day != "Fri"],
 x="total_bill",
 y="tip",
 facet_col="day",
)