Independent test of grid factoring methods.

To address #258, we are making changes to allow 1 values in grid_shape. Previously, we had two methods for computing the factors, I think one slow but simpler, and one fast and more subtle. Only the second one of these is surviving the refactor for this, as the first is effectively unused code. But, the deprecated method being removed might be useful as an independent test of the factorization method that we do use. This deprecated method proved difficult to adapt to allowing the 1 values, which also argued for its removal.

But in case this alternate method is later useful for testing, I note the code used here, so it is not lost track of.

```
def test_both_methods(self):
"""
Do the two methods of computing the multiplicative partitions agree?
"""
for s in [2, 3]:
for n in range(2, 512):
self.assertEqual(utils.mult_partitions(n, s),
utils.create_factors(n, s))

def divisors(n):
i = 2
while i>> mult_partitions(52,3)
[(2, 2, 13)]
>>> mult_partitions(52,2)
[(2, 26), (4, 13)]
"""
return [tuple(flatten(p)) for p in mult_partitions_recurs(n,s)]

def mult_partitions_recurs(n, s, pd=1):
if s == 1:
return [n]
divs = divisors_minmax(n, pd, int(sqrt(n)))
fs = []
for d in divs:
fs.extend([(d,f) for f in mult_partitions_recurs(n/d, s-1, pd)])
pd = d
return fs
```

Start by locating the current multiplicative-partition implementation and checking whether the deprecated create_factors helpers are still present after the grid_shape refactor. Compare the two approaches described in the issue, then determine whether an independent test can cover the surviving method, including grid shapes containing 1 values; done means the test is useful and passes for the supported cases.

Written by the indexing model from the issue text.