Quotient ring with old_poly_ring behaves incorrectly

[Jex-y]

When constructing a quotient ring, sometimes terms that should be dropped by the quotient are left in e.g.:

>>> from sympy import *
>>> d1, d2, d3 = symbols('d_1 d_2 d_3')
>>> InfRing = RR.old_poly_ring(d1,d2) / [d1**2, d2**2]
>>> d_1 = InfRing.convert(d1); d_2 = InfRing.convert(d2)
>>> print((d_2+d_1)**2)
2.0*d_1*d_2 + 1.0*d_2**2 + <1.0*d_1**2,1.0*d_2**2>

The output of this should be:

2.0*d_1*d_2 + <1.0*d_1**2,1.0*d_2**2>

This is not always the case as the following statements are correct:

>>> print(d_1**2)
0 + <1.0*d_1**2,1.0*d_2**2>
>>> print(d_2**2)
0 + <1.0*d_1**2,1.0*d_2**2>
>>> print((d_1 + 1) **2)
2.0*d_1 + 1.0 + <1.0*d_1**2,1.0*d_2**2>

[jksuom]

2.0*d_1*d_2 + 1.0*d_2**2 is one of the weak normal forms of the result. It is not incorrect. It is computed by sdm_nf_mora in distributedmodules. From the docstring:

    Weak normal forms are defined in [SCA, defn 2.3.3]. They are not unique.
    This function deterministically computes a weak normal form, depending on
    the order of `G`.

There are other normal form functions, sdm_nf_buchberger and sdm_groebner, also implemented in distributedmodules but they are not currently applied in agca.modules, presumably for efficiency.