bug fix in variable density implementation

[jzmejia]

fix to Crevasse.paterson_empirical_relation(), added check to accepted depth array to ensure positive values.

This commit closes #112

A new problem has emerged in executing IceBlock.increment_time() in Crevasse.evolve_crevasses()

ib=IceBlock(1140,
            1, 
            dt=1, 
            years_to_run=2,
            thermal_freq=365,
            crev_spacing=56,
            u_surf=200,
            T_profile=pd.read_csv('crevprop/data/foxxTemperatureProfile.csv',
            names=['t','z']),
            sigmaT0=128e3,
            creep_table=df,
            include_creep=True,
            Qin_annual=1000,
            shear_modulus=1e8,
            initial_crev_depth=40.2
           )

crevasse = ib.crev_field.crev_instances[0]

ib.increment_time()

Produces the following Error Messages

Error resolved for Qin_annual = 5000 and a crevasse evolution as follows is produced.
The issue here is that water height goes to 0 and remains there on day 22 and afterward.