GH-16592 Fix NPE when scoring CoxPH mojo from h2o 3.32.x.x

[Mathanraj-Sharma]

Fixes #16592

CoxPH mojo models generated using 3.32.x.x throws NPE when using with h2o >= 3.42.0.4

stacktrace: 
java.lang.NullPointerException
	at hex.genmodel.algos.coxph.CoxPHMojoReader.readModelData(CoxPHMojoReader.java:31)
	at hex.genmodel.ModelMojoReader.readAll(ModelMojoReader.java:260)
	at hex.genmodel.ModelMojoReader.readFrom(ModelMojoReader.java:66)
	at hex.generic.Generic$MojoDelegatingModelDriver.importMojo(Generic.java:122)
	at hex.generic.Generic$MojoDelegatingModelDriver.computeImpl(Generic.java:107)
	at hex.ModelBuilder$Driver.compute2(ModelBuilder.java:253)
	at hex.generic.Generic$MojoDelegatingModelDriver.compute2(Generic.java:92)
	at water.H2O$H2OCountedCompleter.compute(H2O.java:1689)
	at jsr166y.CountedCompleter.exec(CountedCompleter.java:468)
	at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263)
	at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:976)
	at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1479)
	at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104)

The issue mainly occurs since num_numerical_columns and num_offsets are absent in CoxPH mojo from older h2o3 versions.

I am using the -1 as default for num_numerical_columns ( assuming number of numerical columns won't be less than 0 in any cases) and scoring based on it.

ref: https://support.h2o.ai/a/tickets/110207

[valenad1]

I am using the -1 as default for num_numerical_columns ( assuming number of numerical columns won't be less than 0 in any cases) and scoring based on it.

Hi @Mathanraj-Sharma , just realized.. what if we traing CoxPH model only with categorical columns?

[Mathanraj-Sharma]

@valenad1 fix works for model trained using categorical features only as well

[valenad1]

Thank you!