Variable density for firn layer enhancement

[jzmejia]

Issue identified after including depth dependent density functionality to Crevasse class, features added in commits: #111 #110

Crevasse evolution with variable density is producing crevasse shapes like below where they are very bloated at the base, narrowing towards the surface. These crevasses are too wide at the base

crevasse width (m) for 51 model runs

Is the problem in the d equation used to calculate wall displacement not able to support densities other than that of solid ice?

[jzmejia]

Reopening first issue was resolved but a larger problem identified for unstable solutions (mathematical problem)

Depth-variable formation for the second component of the stress intensity factor KI(2) is invalid/unstable for large depths. Need to change implementation/reformulate equations

IMPORTANT: Problem in mathematics applied for solving linear elastic fracture mechanics, not in the code

[jzmejia]

comparison between stable formulation (van der veen 2002) vs unstable formulation at large depths (van der veen 1998)

[jzmejia]

Fix to problem:

1. Find intersection between constant density and variable density KI(2) solutions.
   Crevasse.threshold
2. Use Crevasse.threshold to parse KI(2) solutions such that at depths less than threshold the variable density solution is used and at values greater than or equal to threshold the constant density solution is used.

[jzmejia]

density was manually changed inside of some function and density was filled with 917 below a value. Find this and remove it so that the crevasse wall doesn't have a little malformity at that depth.