[FEA] Number of threads should be configurable in CPU nvForest

[hcho3]

Is your feature request related to a problem? Please describe.
Currently, the CPU nvForest uses the default number of OpenMP threads:

nvforest/cpp/include/nvforest/detail/infer_kernel/cpu.hpp

Lines 100 to 104 in 6e23cc9

	#pragma omp parallel num_threads(std::min(index_type(omp_get_max_threads()), task_count))
	{
	// Infer on each grove and chunk
	#pragma omp for
	for (auto task_index = index_type{}; task_index < task_count; ++task_index) {

Describe the solution you'd like
Users should be able to configure the number of CPU threads.