LIM2 VP/EVP initialisation and allocation corrections

[acc]

A couple of minor corrections are needed to the LIM2 code. The first fixes an allocation problem in iceini_2.F90 when using key_lim2_vp:

--- NEMOGCM/NEMO/LIM_SRC_2/iceini_2.F90 (revision 2853)
+++ NEMOGCM/NEMO/LIM_SRC_2/iceini_2.F90 (working copy)
@@ -22,7 +22,8 @@
    USE dom_ice_2        ! LIM2 ice domain
    USE par_ice_2        ! LIM2 parameters
    USE thd_ice_2        ! LIM2 thermodynamical variables
-   USE limrhg           ! LIM2 rheology
+   USE limrhg           ! LIM2 EVP rheology
+   USE limrhg_2         ! LIM2  VP rheology (requires "key_lim2_vp")
    USE ice_2            ! LIM2 ice variable
    USE limmsh_2         ! LIM2 mesh
    USE limistate_2      ! LIM2 initial state
@@ -64,6 +65,8 @@
       ierr = ierr + thd_ice_alloc_2()       ! thermodynamics
 #if ! defined key_lim2_vp
       ierr = ierr + lim_rhg_alloc  ()
+#else
+      ierr = ierr + lim_rhg_alloc_2  ()
 #endif

The second initialises stress1_i, stress_2_i and stress12_i correctly when starting from stratch (& using EVP rheology). Failure to do so leads to NaNs? on some systems/configurations:

--- NEMOGCM/NEMO/LIM_SRC_2/limistate_2.F90      (revision 2853)
+++ NEMOGCM/NEMO/LIM_SRC_2/limistate_2.F90      (working copy)
@@ -123,6 +123,11 @@
       sxxst (:,:)  = 0.e0
       syyst (:,:)  = 0.e0
       sxyst (:,:)  = 0.e0
+#if ! defined key_lim2_vp
+      stress1_i (:,:) = 0._wp                          ! EVP rheology
+      stress2_i (:,:) = 0._wp
+      stress12_i(:,:) = 0._wp
+#endif

These changes have been made to the trunk at revision 2855