Evaluation of layer up to which LTE should be computed not adapted to parallel case

[emillour]

When computing up to which layer LTE computations should be done in nlthermeq.F , it is assumed in the code that there is no parallelism...
So in practice this is wrong when running the GCM in parallel. Fortunately in hybrid coordinates the designated pressure up to which LTE computations are to be done (pminte=0.04Pa) is actually in the "purely pressure" region of the vertical coordinate and thus insensitive to location or surface pressure.

But this is not clean. nlthermeq.F should be re-written and explicitly handle the parallel case, making sure results do not differ from when not in parallel.