Update min/max value of remote datasets for recent upstream changes

[seisman]

Description of proposed changes

See #2961 (comment) for context.

The min/max values of remote datasets are out of our control so it makes little sense to check them in PyGMT tests. This PR remove all checks about min/max values of remote datasets, with one exception earth_mask, since we know that the earth_mask dataset has discrete values from 0 to 4.

[weiji14]

Ok with removing the min/max value checks, but we should still have some check to ensure that the pixel values returned are valid. Maybe something like a notnull() check, or some other statistical measure?

[seisman]

Maybe something like a notnull() check,

notnull() checks make no sense to some datasets (e.g., earth_age), which contains NaN values.

or some other statistical measure?

It means we still need to update these statistical values when the datasets change.

[seisman]

I've added the checks of min/max values back to the tests. It means more maintenance burden but at least the values can make sure that the loaded data have reasonable values.