Currently, the numerical queries cf.le, cf.ge and cf.wi are susceptible to rounding issues, as the inequality operators can not be modified by the settings of rtol and atol.

This could be changed by replacing the operations with a strict inequality operation and a numerically tolerant equality operation:

E.g.

cf.le(5) = 5.00000000000001

would effectively be implemented as

5 < 5.00000000000001 | numpy.allclose(5, 5.00000000000001, rtol=rtol, atol=atol)

This would help greatly when using cf.Query in, for example, subspace definitions.

API: the functions should have rtol and atol keyword parameters. (cf.wi(3, 4, rtol=45, atol=0.0002))

Possible issue: Under this scheme, cf.wo would not be the exact inverse of cf.wi