Custom weights for DBStream

[th3sh3ph3rd]

Currently, river's DBStream implementation sets the weight of every new microcluster to 1. This is also in accordance with the algorithm proposed in the original paper.

However, it would be very useful to set custom weights when calling learn_one(), as it would allow assigning varying importance to newly incoming data points. Furthermore, the learn_one() interface already includes an unused parameter w, which could be used for this purpose.

I propose to add this functionality to DBStream. Only minimal changes, concretely in lines 179, 187, 191, 205, 225 and 402, would be necessary to add great value and flexibility to the DBStream algorithm.

If it is decided that this addition is useful, I would also be open to contributing these changes to the river project.

[MaxHalford]

As mentioned in your other post, feel free to propose an implementation 👍

[th3sh3ph3rd]

Duplicate due tu github problems. Further discussion in #1716.