KPP now supports LMD computation of turbulent scales

By default, CVMix continues to use the Danabasoglu, et al., formulation for computing turbulent velocity scales (where sigma is capped at some epsilon (0.1 by default) for both stable and unstable regimes... but an option now exists to use the Large, et al., formulation (where sigma is only capped for unstable conditions).

See paragraph preceding Eq (13) in LMD94 and paragraph preceding Eq (A3) in DLTGBM06 (doesn't quite roll of the tongue as easily as LMD94...) for more details.