Misorder labels from contructed tensor after tensor factorization

Hey @cameronmartino, @gwarmstrong,
Thanks for this awesome tool.

I was concerned about the sorting procedure for loadings when fitting tensor factorization, would it cause misorder of loading labels in the `label` step?

_Codes of sorting procedure excerpted from `factorization.py` line 219 - 224 in `_fit` function_

```
# save array of loadings for subjects
self.subjects = loads[0].copy()
self.subjects = self.subjects[self.subjects[:, 0].argsort()]
# save array of loadings for features
self.features = loads[1].copy()
self.features = self.features[self.features[:, 0].argsort()]
```
_Codes of labeling step excerpted from `factorization.py` since line 335 in `label` function_

```
# DataFrame single non-condition dependent loadings
self.subjects = pd.DataFrame(self.subjects,
columns=self.biplot_labels,
index=construct.subject_order) # self.subjects reordered, but construct.subject_order didn't
self.features = pd.DataFrame(self.features,
columns=self.biplot_labels,
index=construct.feature_order) # self.features reordered, but construct.feature_order didn't
......
```

Start in factorization.py at _fit around lines 219–224 and then follow the label function from line 335. Compare the sorting of self.subjects and self.features with construct.subject_order and construct.feature_order, using a tensor-factorization example to inspect the resulting labels. Done means labels remain aligned with their loading rows after factorization and labeling.

Written by the indexing model from the issue text.