Add numtest and use doctest.ELLIPSIS_MARKER to fix doctest errors when doing floating point number comparisons

[justaddcoffee]

doctests in survival.rst and phenotype_scores.rst are failing for me on my Macbook M1 MacOS v12.5

FAILED               [  2%]
docs/user-guide/analyses/survival.rst:0 ([doctest] survival.rst)
120 
121 We execute the analysis by running
122 
123 >>> result = survival_analysis.compare_genotype_vs_survival(
124 ...     cohort=cohort,
125 ...     gt_clf=gt_clf,
126 ...     endpoint=endpoint,
127 ... )
128 
129 >>> result.pval
Expected:
    0.06200425830044376
Got:
    0.06200425830044377

/Users/jtr4v/PythonProject/gpsea/docs/user-guide/analyses/survival.rst:129: DocTestFailure

This PR adds numtest as a test dependency and then uses ELLIPSIS_MARKER to allow approximate comparisons

[ielis]

Hello @justaddcoffee,

thanks a lot, using ellipsis in float comparisons is a very good and useful idea. I have a few comments & questions, can you pls see below?

Thank you!

[ielis]

The doctest runner seems to recognize the ... even without #doctest: +ELLIPSIS marker, and the test passes here. Does the same happen on your end? If yes, pls remove # doctest: +ELLIPSIS.

[ielis]

I am not sure we really need this dependency. All tests passed on my end without numtest, just with the ... at proper places. Unless I am missing something, we do not need numtest, and by applying Occam's razor, it should be removed.

[justaddcoffee]

Thanks @ielis - you are right, the numtest dependency, doctest.ELLIPSIS_MARKER = '...' and DOCTEST +ELLIPSIS annotations are not necessary here, so I've removed them. It looks like all this ELLIPSIS stuff is now default, and I was reading some stale documentation

[ielis]

Looks great now, thank you!