Add keyword arg guidelines to contrib doc

[sirosen]

In response to some discussion around **kwargs, we agreed that I should take a first crack at these guidelines.

I tried to provide some context without lengthy exposition, and I've included a link to the PEP which introduced the * marker.
One thing I debated, but decided against, was any note about "new functions" (vs in general) and the theme that these rules are broken not-infrequently for backwards compatibility. I think that's a given in any codebase with this level of history, so I opted to just make the blanket statements and trust that the compatibility concerns are implicitly clear.

Backwards compatibility around **kwargs capturing

I avoided explaining in the text of the contrib doc, but, to illuminate the mention of "a class of backwards compatibility concerns", here's a bit of detail.

There are cases in which adding keyword arguments is not incompatible or not majorly so. This is completely compatible:

# v1
def foo(**kwargs):
    return {k: v for k, v in kwargs.items() if v is not None}

# v2
def foo(bar: str | None = None, **kwargs):
    data = kwargs.copy()
    if bar is not None:
        data["bar"] = bar
    return {k: v for k, v in data.items() if v is not None}

But there are simple cases in which it becomes nasty quickly. Here's a trivial-if-silly example:

def foo(a: int | None = None, b: int | None = None, **kwargs):
    data1 = {"a": a, "b": b}
    return (
        {k: v for k, v in data1.items() if v is not None},
        {k: v for k, v in kwargs.items() if v is not None},
    )

Because this function was defined to capture all of **kwargs into a specific area of functionality, it's not possible to add c: int | None = None as an item to include in the data1 component.

Avoiding **kwargs ensures that we don't find ourselves in this sort of unfortunate situation.

---

📚 Documentation preview 📚: https://globus-sdk-python--775.org.readthedocs.build/en/775/