Improve CUDAEnsemble device selection error messages

[ptheywood]

CUDAEnsemble does not provide useful error messages when no CUDA devices are present or the CUDA runtime cannot be created, instead just throwing the error to the use via gpuErrchk

FLAMEGPU2/src/flamegpu/gpu/CUDAEnsemble.cu

Line 75 in f70bac0

gpuErrchk(cudaGetDeviceCount(&ct));

This should be improved to match the more complex CUDASimulation device selection logic that throws exceptions with useful messages.

FLAMEGPU2/src/flamegpu/gpu/CUDASimulation.cu

Lines 1546 to 1575 in f70bac0

	cudaStatus = cudaGetDeviceCount(&device_count);
	if (cudaStatus != cudaSuccess) {
	THROW exception::InvalidCUDAdevice("Error finding CUDA devices! Do you have a CUDA-capable GPU installed?");
	}
	if (device_count == 0) {
	THROW exception::InvalidCUDAdevice("Error no CUDA devices found!");
	}
	// Select device
	if (config.device_id >= device_count) {
	THROW exception::InvalidCUDAdevice("Error setting CUDA device to '%d', only %d available!", config.device_id, device_count);
	}
	if (deviceInitialised !=- 1 && deviceInitialised != config.device_id) {
	THROW exception::InvalidCUDAdevice("Unable to set CUDA device to '%d' after the CUDASimulation has already initialised on device '%d'.", config.device_id, deviceInitialised);
	}
	// Check the compute capability of the device, throw an exception if not valid for the executable.
	if (!util::detail::compute_capability::checkComputeCapability(static_cast<int>(config.device_id))) {
	int min_cc = util::detail::compute_capability::minimumCompiledComputeCapability();
	int cc = util::detail::compute_capability::getComputeCapability(static_cast<int>(config.device_id));
	THROW exception::InvalidCUDAComputeCapability("Error application compiled for CUDA Compute Capability %d and above. Device %u is compute capability %d. Rebuild for SM_%d.", min_cc, config.device_id, cc, cc);
	}
	cudaStatus = cudaSetDevice(static_cast<int>(config.device_id));
	if (cudaStatus != cudaSuccess) {
	THROW exception::InvalidCUDAdevice("Unknown error setting CUDA device to '%d'. (%d available)", config.device_id, device_count);
	}
	// Call cudaFree to initialise the context early
	gpuErrchk(cudaFree(nullptr));