Do not impose mask dtype to be bool

Currently, mask in the SITS class need to be bool (i.e., 0 or 1). In some cases it can be usefull to allow for non-binary mask: e.g., when using probabilistic values of having a cloud/shadow.

In the current state of SITSclass, to allow for non binary mask, some modifications are required

• Remove assert in the init function
• In tomethod: handle casting
• In pad_acquisition_time, pass the right type of value (true or 1) to constant_pad_end
• In subset_doy_monomodal_sits : check actual mask dtype for init out_mask tensor
• In sits_where: handle the logical_and properly

Futhermore, others methods based on SITS assume mask dtype is boolean (e.g., mutan and linear_gapfilling) and does not check it (delegates it to SITS constructor). I am wondering what is the best solution to allow non binary initialization: add a flag that desactivate the assert ? Something like

class SITS:
    """
    Base class for SITS data
    """

    def __init__(
        self, data: torch.Tensor, doy: torch.Tensor, mask: None | torch.Tensor = None, require_boolean_mask: bool=True,
    ):
        assert mask.dtype == torch.bool or require_boolean_mask == False