PPE: Accept more than one distribution as target

[aloctavodia]

Currently PPE, only accepts as a target a single PreliZ distribution. We should allow more than one distribution. I see at least two useful options

• A list, and then we assume all distributions have an equal contribution. i.e. we sample all distributions with the same probability.
• Or a list of tuples, like [(dist0, w0), (dist1, w1)], for unequal contribution. i.e. we sample all distributions proportional to the weight w

To implement this we need to change this line https://github.com/arviz-devs/preliz/blob/main/preliz/internal/optimization.py#L235 into a function that takes target and returns a sample or we can precompute the samples that we pass to the optimization routine (the better option is probably the best one).