This PR makes floats have a total ordering in their comparison operator, where NaNs compare equal to each other (regardless of sign or payload) and greater than any other floating point value. This matches the order used in e.g. duckdb and allows unambiguous sorting, hashing and equality. Ironically, sorting/hashing is not yet included in this PR, just the explicit comparison operators, one step at a time.

Also included in this PR is the new polars-compute crate that is intended to contain pure computational functions that do not require polars-core (but may use polars-arrow). We can start moving other things (such as the efficient summation routine) here after this PR.

Example

Before:

>>> s = pl.Series([1.0, float("nan"), float("inf")])
>>> s == s
shape: (3,)
Series: '' [bool]
[
        true
        false
        true
]

After:

>>> s == s
shape: (3,)
Series: '' [bool]
[
        true
        true
        true
]