Bug with single precision

[semi-h]

#206 added single precision support, but there is a bug in x3d2 when we compile with single precision support. Here is the list of tests that fail:

The following tests FAILED:
	  8 - test_time_integrator_omp_1 (Failed)
	  9 - test_omp_tridiag_omp_4 (Failed)
	 11 - test_omp_transeq_omp_4 (Failed)
	 12 - test_omp_transeq_species_omp_4 (Failed)
	 13 - test_omp_dist_transeq_omp_4 (Failed)
	 19 - test_time_integrator_cuda_1 (Failed)
	 20 - test_fft_cuda_1 (Failed)
	 23 - test_thom_cuda_1 (Failed)
	 25 - test_cuda_transeq_cuda_4 (Failed)

I would start by debugging the time integrator test first, as it only uses vecadd and the vecadd test itself is passing.
Also, the cuda test with the tridiagonal solver passes, and omp test with the thomas algorithm passes, so I guess there is nothing fundamentally wrong with the distd2 algorithm or thomas algorithm with single precision.