Provide fix keyword argument in the middle of pipeline function calls

[balthazarneveu]

This would allows setting a few values in each filter & make some functions more "specific". Right now, in the following example , it's not yet possible to set the style & name used for each histogram curve.

@interactive(bins=(512, [256, 4096, 64]))
def compute_histogram(rgb, bins=512):
    style="-"
    name="histo"
    luma = torch.mean(rgb, dim=(-3))
    # histo, histo_bins = torch.histogram(luma, bins=256, range=(0,1), density=True) # Not working on a GPU
    histo, histo_bins = histogram(luma, bins=bins, range=(0,1), density=True)
    return (histo, histo_bins, style, name),

# ----------------------------------------------------------------------------
def stack_histograms(*histos) -> torch.FloatTensor:
    """Stack several histograms into a curve
    No custom legend so far!
    """
    histo_curve = Curve(
        [
            (histo_bins[:-1].detach().to("cpu").numpy(), histo.detach().to("cpu").numpy(), "-", f"{name} {index}")
            for index, (histo, histo_bins, style, name) in enumerate(histos)
        ],
        grid=True,
        title="Histogram",
        xlim=(0., 1.),
        ylim=(0., 0.01)
    )
    return histo_curve

stack histograms works fine (no need to provide the exact amount of inputs . BUT you can't set style or names in the main pipeline . I wished I could write: hist = compute_histogram(rgb , style="b-", name="rgb gamma")