[Breaking] Remove backend generics from high-level API

[laggui]

Now that we have a Dispatch backend (enum static dispatch) we can remove the backend generics from the high-level Tensor API to use this as the default backend.

Example module:

#[derive(Module, Debug)]
pub struct BatchNorm {
    /// The learnable weight gamma.
    pub gamma: Param<Tensor<1>>,
    /// The learnable weight beta.
    pub beta: Param<Tensor<1>>,
    /// The running mean.
    pub running_mean: RunningState<Tensor<1>>,
    /// The running variance.
    pub running_var: RunningState<Tensor<1>>,
    /// Momentum used to update the metrics.
    pub momentum: f64,
    /// A value required for numerical stability.
    pub epsilon: f64,
}

Although a big breaking change, it should remove a lot of friction with the current API, and allow for easy runtime backend selection.

This also means a CPU device can always be available (unless disabled), making host-side tensor manipulations easily accessible. Beforehand, one would have to explicitly enable another backend (e.g. ndarray) and move the data between the "primary" backend -> tensor data -> ndarray which was annoying.

This change should also bring a new high-level device struct to handle the default precision settings and device operations (e.g. device.sync()).