Fix OTT-JAX tests

[emdupre]

From run on 31 July. Relates to #53.

=========================== short test summary info ============================
FAILED fmralign/tests/test_alignment_methods.py::test_ott_backend - AssertionError: 
Arrays are not almost equal to 3 decimals

Mismatched elements: 95 / 400 (23.8%)
Max absolute difference: 0.91489681
Max relative difference: 0.95004539
 x: array([[2.268e-03, 6.906e-04, 1.016e-04, 3.824e-07, 1.470e-03, 1.189e-09,
        2.241e-07, 1.104e-04, 1.422e-06, 9.785e-09, 6.776e-09, 4.613e-06,
        2.108e-02, 3.041e-03, 3.216e-06, 1.100e-02, 5.124e-04, 1.619e-07,...
 y: array([[4.537e-02, 1.381e-02, 2.033e-03, 7.641e-06, 2.939e-02, 2.378e-08,
        4.483e-06, 2.209e-03, 2.844e-05, 1.957e-07, 1.355e-07, 9.224e-05,
        4.216e-01, 6.082e-02, 6.432e-05, 2.201e-01, 1.025e-02, 3.238e-06,...
FAILED fmralign/tests/test_pairwise_alignment.py::test_models_against_identity - assert 0.010373740716492528 >= 0.34642806848096425
FAILED fmralign/tests/test_template_alignment.py::test_template_closer_to_target - assert 0.05907143983986113 >= 0.8538914990411407
============ 3 failed, 12 passed, 797 warnings in 115.25s (0:01:55) ============

Exited with code exit status 1

[emdupre]

@bthirion I hadn't kept this line, which was causing the issues :

P.matrix * len(X.T)

I'm not sure that I understand, though, why we want to multiply by the length of the time series (since this isn't duplicating the matrix but multiplying each value in-place ? If you have a reference you'd recommend, I'd appreciate any pointers !

[emdupre]

@bthirion I'll merge the associated PR but keep this issue open just so we don't lose track of this q !

[emdupre]

Closing with #72