Add UniformBinning to schemav2

[eguiraud]

When defining a Binning, the edges attribute can now be either a list of floats (like before) or an instance of UniformBinning(n, low, high). When using the latter option, the Binning correction is optimized to use a simple arithmetic formula rather than binary search to look up the bin index.

This is a fresh (better) take on #181 .

Example usage:

from correctionlib import schemav2 as schema

if __name__ == "__main__":
    corr = schema.Correction(
        name="test",
        version=2,
        inputs=[schema.Variable(name="x", type="real")],
        output=schema.Variable(name="a scale", type="real"),
        data=schema.Binning(
            nodetype="binning",
            input="x",
            edges=schema.UniformBinning(n=2, low=0., high=2.),
            content=[1.0, 2.0],
            flow="error",
        ),
    )
    e = corr.to_evaluator()
    print(e.evaluate(0.5))
    print(e.evaluate(1.0))
    print(e.evaluate(3.0)) # errors out

To do:

• add tests
• add support for MultiBinning

[nsmith-]

Yeah this is the right approach I think.
Looks good, some small extra validation comments only

[eguiraud]

Addressed review comments, added some tests for Binning+UniformBinning, fixed a bug in an edge case that the test found.

I will work on adding support for MultiBinning as well as soon as possible (could also be a separate PR).

[nsmith-]

I'm fine to wait for MultiBinning in this PR

[nsmith-]

Looks great! Thanks so much for this addition, and sorry I didn't review it sooner