Fix issues with is_jacobian_term_used()

[maximumcats]

There were three problems here. First, the logic within the function was incorrect: a Jacobian term is used if both spec1 and spec2 are connected to the same rate, rather than if either species is used. Second, this function is indexing the species using 1-indexing, but we were calling it with zero-indexed species inside the linear algebra. Third, we can only skip entries if we're using the analytical Jacobian. The numerical Jacobian will have some additional non-zero entries since it captures some physics, such as the screening, which we are not currently able to conclude in the analytical netowrk implementation.

[zingale]

closes #1499

[zingale]

these changes cause test_linear_algebra to fail:

original x and x from the solve (RHS:: solve): 
                   1    7.162462030968882
                  20    13.97277986846644
                   3   -0.822910119156765
                  40    33.89190438198373
                   5    2.117609026004514
                  60    56.02163984189685
                   7    3.458108032248415
                  80    75.97253851812479
                   9    6.080893721649049
                 100    96.56668190864372
                  11    8.490387506470494
                 120    117.0183692563761
                  13    10.03295745366218
                 140    140.7876345006217

[zingale]

we worked around this for now in #1502