Corrected a bug in the correlation and normalization driver

[arango]

Correted a problem in the correlation.h driver that was forcing to compute the error covariance normalization coefficient for initial conditions, model error, open boundary conditons, and surface forcing at the same time.

It was not possible to compute one set of coefficients alone and we were getting I/O error because trying to process or access non-existing normalization and standard deviation files.

I also added special logic in inp_par.F when the CORRELATION CPP options is activates.

Many thanks to Rich Signell for bringing this to my attention.