NaN or Inf values in cold gas fractions plots

[EvgeniiChaikin]

The data in several cold-gas-fractions plots may have NaN values (depending on the run and redshift), which makes the pipeline crash in certain cases.

E.g. sigma_h2_to_neutral_H_fraction_100

Possible solution: always use mask whenever there is a change to have division by zero.

[JBorrow]

Yes, I've been concerned about this previously.

We do not want to just mask when there is a division by zero. That's the wrong thing to do here. There will be a physical reason that there is no HI, and in the real universe we wouldn't be able to measure the HI in those galaxies - so we should cut by that physical reason (e.g. just plot for star forming galaxies in a given mass range).