[cuda] testEigenGPUNoFit test fails

[makortel]

This issue is a follow-up to #40.

Instructions to reproduce

$ make cuda
$ ./test/cuda/testEigenGPUNoFit
TEST EIGENVALUES
TEST INVERSE 3x3
TEST INVERSE 4x4
TEST INVERSE 5x5
terminate called after throwing an instance of 'std::runtime_error'
  what():  
.../pixeltrack-standalone/src/cuda/test/testEigenGPUNoFit.cu, line 170:
cudaCheck(cudaDeviceSynchronize());
cudaErrorLaunchFailure: unspecified launch failure

Aborted (core dumped)

[makortel]

cuda-memcheck gives

========= Unknown Error
=========
========= Program hit cudaErrorLaunchFailure (error 719) due to "unspecified launch failure" on CUDA API call to cudaDeviceSynchronize. 
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/lib64/libcuda.so.1 [0x3c43a3]
=========     Host Frame:.../cmssw/slc7_amd64_gcc820/external/cuda/10.2.89/lib64/libcudart.so.10.2 (cudaDeviceSynchronize + 0x166) [0x4fb26]
=========     Host Frame:./test/cuda/testEigenGPUNoFit (_Z14testInverse5x5v + 0x730) [0x83f0]
=========     Host Frame:./test/cuda/testEigenGPUNoFit (main + 0x18) [0x4358]
=========     Host Frame:/lib64/libc.so.6 (__libc_start_main + 0xf5) [0x22505]
=========     Host Frame:./test/cuda/testEigenGPUNoFit [0x4437]
=========
terminate called after throwing an instance of 'std::runtime_error'
  what():  
.../pixeltrack-standalone/src/cuda/test/testEigenGPUNoFit.cu, line 170:
cudaCheck(cudaDeviceSynchronize());
cudaErrorLaunchFailure: unspecified launch failure

========= Error: process didn't terminate successfully
========= No CUDA-MEMCHECK results found

[makortel]

By random poking the assignment in

pixeltrack-standalone/src/cuda/test/testEigenGPUNoFit.cu

Line 33 in 4929b53

__global__ void kernelInverse5x5(Matrix5d *in, Matrix5d *out) { (*out) = in->inverse(); }

is the problem (the call to inverse() works, unless it gets optimized away by the compiler).