New testing

[pedrobcst]

This PR does the following:

Implements new testing method. Currently we are testing BOTH simulations of crystal structures and REFINEMENT against a dummy data (default HoB2.1). After changing how the refinement is done by default, and by changing in the queriers (limiting the volume of the crystal structures coming frm COD, etc), Xerus became much stabler. Furthermore, we are catching problematic simulations in place instead now, so the simulation test is not necessary anymore. However, since GSASII uses it own CIF parser, some tests are still required. Therefore, the new testing is only the following:

• Tests with GSASII can successfuly read a CIF coming from a given provider. There are few structures that GSASII fails to parse, so this is necessary.

Also:
We reran the new testing (togethrer with the new OPTIMADE API that gives more structures from OQMD) and we found that:

• The addition of more crystal structures increased the accuracy in the bench from 84% to 88/92%.
• Still have issue when dealing with distorted low-temp structures for a same chemical composition

Note: 92% is if you accept that a '3' phase sample actually got matched with 4 phases in automatic (3 correct + one extra).

Closes #23