Support running an algorithm multiple times with the same parameter combination

[agitter]

Omics Integrator 1 is sometimes run with the same parameter combination many times when r is nonzero, which introduces edge noise. That is not currently possible in SPRAS.

We could add a new parameter type in the config file that is the number of times to run with that parameter combination. The config parser would expand that into a unique integer counter.

This could be useful for any non-deterministic algorithm.

[tristan-f-r]

(This was also asked in #33, though this issue could probably supersede #33 after #228 is merged)

[ntalluri]

This was brought up at my thesis committee as well. For the registered report/paper, we should prioritize adding this feature into SPRAS.

[tristan-f-r]

Since this wouldn't conflict with #228, I can work on it.

I'll add a parameter repeat in params.

[ntalluri]

@tristan-f-r please wait on doing this issue. I need to think through the issue.

[agitter]

We should brainstorm the ideal solution. I envisioned we should have some new integer parameter, let's called it repeats for now, that runs each unique parameter combination that many times. Internally SPRAS would include a counter in the parameter combination dictionary so that we would still get unique parameter combinations and hashes. That would probably let us continue ensembling outputs in the same way we do now, but that downstream analysis is what requires some consideration before we move ahead.

We also need to avoid conflicts between any new internal parameter like counter and repeats and real parameters from the supported algorithm. That could be handled by making the internal parameter spras_counter, for example.

[tristan-f-r]

In #335, one could use different values for the algorithm-specified seed parameter (or the fake _seed parameter) instead, and we can have a utility function to automatically generate n seeds as well.