Add mechanism to constrain fitting parameters

[anloes]

Sometimes, neutcurve generates a curve for a given replicate with a slope or top that is aberrant with other replicates, potentially due to noisy data. This may be able to be resolved by adding a mechanism to constrain the fitting parameters to a reasonable slope or top, we could add a mechanism to do this to the package.

Attached (and shown in plots included) are several examples of curves where this issue was observed.
240104_curveswithdifferentslopes.csv

[anloes]

@jbloom and @ckikawa this issue is similar to what we were discussing for the data with the new libraries.

[jbloom]

Yes, I agree. I am hoping @ckikawa can add her relevant data here so I have more test cases for this, and whenever that is done I am going to start working on this.

[ckikawa]

@jbloom, here are some test cases from the H3N2 library. I've separated these into cases where replicates are called widely varying tops, usually in the case where virus-barcodes start at 0 fraction infectivity and are unable to consistently infer accurate tops:

240322_curves_with_different_tops_fits.csv
240322_curves_with_different_tops_frac_infectivity.csv

And other cases where replicates have widely varying slopes, usually in the case where neutralization takes place over most of the entire window of serum dilutions:

240322_curves_with_different_slopes_fits.csv
240322_curves_with_different_slopes_frac_infectivity.csv

[jbloom]

@ckikawa @anloes, this shows the data Caroline posted with constraints on the fit parameters:
https://jbloomlab.github.io/neutcurve/constrain_params_range.html

I will work on integrating this into seqneut-pipeline.