Make `plot_contour` example more pedagogical

Motivation

Examples in the API reference should be pedagogical. The plot for the contour plot is not with somewhat poorly visible categorical parameter values.

Description

The contour plot example could use a better suited objective function e.g. limited to floats. The following code was provided by @toshihikoyanase . Maybe we can modify it to only suggest 2 variables to keep it simple and use it here.

def objective(trial):
    x = trial.suggest_float("x", 0, 1)
    y = trial.suggest_float("y", 0, 1)
    z = trial.suggest_float("z", 0, 1)
    return (x - 0.5) ** 2 + (y - 0.2) ** 2 + (z - 0.7) ** 2

1. Update example code.
2. Update example image generation script.
3. Run the script from 2. to generate a new image.

[hvy]

Addressing the visibility for categorical samples in the first place might be an issue in itself but this is probably less straight forward since we cannot compute distances between categorical values.

Edit: this is an issue for suggest_int values as well. We might want to consider creating some padding after all.

[toshihikoyanase]

Thank you @hvy, and let me share my notebook that draws the contour plot using the above objective function.