Minor improvements to parallelization

[pmocz]

Removing some unused variables in parallel loops, and improving a couple omp pragma statements for tiny speedup.

To make future progress in parallelizing Mesa, we will need to parallelize the bcyclic solves when the # threads > # blocks