fix(test_from_matrix): assume max precision

[MarcoMiretti]

When converting to matrix back and forth, coordinates appear to be
rounded to a precision of 1e-8. This causes sporadic failures in said
test.

For this test it should be assumed that the coordinates of arr can be
described with said precision.

Solves issue #30

Example of what i think is causing the problem:

>>> q = Quaternion(*arr)
>>> q
Quaternion(0.9694471267866732, 0.0, 8.711074176117386e-16, 0.2453003635648832)
>>> m = q.matrix
>>> m
array([[ 8.79655463e-01,  4.75611465e-01, -1.68898517e-15],
       [-4.75611465e-01,  8.79655463e-01,  4.27365932e-16],
       [ 1.68898517e-15,  4.27365932e-16,  1.00000000e+00]])
>>> q.from_matrix(m)
Quaternion(0.9694471267866732, 0.0, 1.0536712127723509e-08, 0.24530036356488336)

[MarcoMiretti]

Hmm. i guess something funny is happening with precision in python 3.4 🤔

[coveralls]

Coverage remained the same at 98.467% when pulling b911ed2 on MarcoMiretti:fix-test-from-matrix into 649d47b on satellogic:master.

[MarcoMiretti]

Hmm. i guess something funny is happening with precision in python 3.4 thinking

Fixed it changing the approach. Instead of assuming that the values are rounded, just round them at the beginning.