Add invariance tests

[arthur-lin1027]

For the pairwise expansion, I don't see the transformation over rotations via the wigner d matrix explicitly implemented within the code (eqn 7 in the preprint). Is this something that's implicitly being done? Eqn 7 is reproduced below:

$$\rho_{nlm} = \sum_{m'=-l}^l D^l_{mm'}(R) \int_{\mathbb{R}^3} d^3(\boldsymbol{r'})~g({R\left(\boldsymbol{r'} - \boldsymbol{r_{ij}}'\right)} R_{nl}{r'} Y_{lm'}({\hat{r}'})$$

For sanity we should add some tests to test_ellipsoidal_density_projection.py to ensure that rotational invariance is established, both for the pairwise coefficients and the three-body expansion.

[curiosity54]

This is not explicitly done in the code, but the rho_nlm coefficients being used have been derived this way. A good way to test this is to compare the coefficients of structure A and B, where B is a rotated copy of A (lets say the rotation is R). For each l, the coefficients must rotate as rho_{nlm}(B) = \sum_{m'} D^l_mm'(R) rho_{nlm} (A). This would ensure, for instance, rho_{n 00}(B) = rho_{n00}(A)

[arthur-lin1027]

While not a explicit test per-se, the @SeonwooH made a notebook that was incorporated into #50 (PR: #54 ) , demonstrating the invariances. I'll keep this branch alive in case we want to take what he did in the notebook and formally integrate it into a test.