TST: Check dataset classes' optional parameters' integration

[jhlegarreta]

Test dataset classes' optional parameters' integration with estimator.

[jhlegarreta]

This PR was motivated by issue #223. It may seem extremely verbose and an overkill to test all this for the estimator, obut the estimator has a careful design that has some complexity and does a lot of things that seem hard to be outsourced. I am happy to transfer this to another testing module if necessary.

Although not apparent in the CI results due to a first error execution halt, running them locally surfaces a design problem related to the use of b0s: the DKI model requires multi-shell data and DIPY's internal check considers the b0s for the computation. As NiFreeze masks b0s, and thus, passing b2000 and b3000 is not considered enough.

So, related to issue #79 and PR #82.

The below check, which reproduces the DIPY logic, is put in the test so that the test can safely skip the single-shell cases:
https://github.com/nipreps/nifreeze/pull/313/files#diff-3f12d72fb4f7251c22997c345d4866aa01b84aef3c94298248a94749515ff6f3R412-R419

I can split the DKI test into a separate PR until somebody takes the time to address this, and in order to move forward with the rest.