Pass kwargs to fit in Workflow

[LouisCarpentier42]

There is an option to pass kwargs to the .fit() method of the anomaly detectors. These kwargs are passed to the compute_window_size() method, which can automatically compute the window size for a given time series. These arguments are not used within a workflow, and thus you can not adapt them. It would be useful to add some way to access these parameters from within the Workflow.

[LouisCarpentier42]

There might be two additional features that are interesting with this:

1. It would be nice if you could adapt the kwargs for each anomaly detector (or each type of anomaly detector) independently. For example, for IForest use default_window_size=64, but for Matrix Profile use default_window_size=48.
2. Further, it would be interesting to be able to set the parameters based on the time series. Currently, the upper and lower bound on the window size are fixed integers. It would be interesting to be able to pass a float in [0, 1] which expresses a percentage. For example, if upper_bound=0.5, then the upper limit would be 500 for a time series of length 1000, and 50 for a time series of length 100. A problem with this could be that, if the time series is very long (e.g., 100.000 observations), then computing a window size could take a very long time. So a combination of the two (integer bound and float bound) would be even better.