Improvements to `extreme_values_analysis`

[RondeauG]

Addressing a Problem?

Using this module for the first time, there are a few things that I think that we should improve:

• For the confidence interval, there should a choice between returning 3 variables (as we currently do) and combining them under a new dimension quantile (e.g. [0.05, 0.50, 0.95]).
• We should implement a way to compute multiple return levels at once, since in its current form, we'd need to completely recompute the fit every time.
• Related to this, we should find a way to use the parameters computed using fit to compute return levels, even if that's not possible with Extremes.jl. Can we simply use xhfa.local.parametric_quantiles? Would that work with the confidence intervals?

Additional context

@ospinajulian

Contribution

• I would be willing/able to open a Pull Request to contribute this feature.