Optimization of data initialization for large sparce datasets #16053

Job	Run time
Build CUDA + manylinux_2_28_x86_64 wheel	12m 3s
Build CPU	14m 15s
Build GPU-enabled R package	23m 47s
Build CUDA with RMM	12m 35s
Build CUDA with RMM (dev)	12m 57s
Build CPU wheel for manylinux2014_x86_64	2m 18s
Build manylinux_2_28_aarch64 wheel	3m 35s
Build CPU wheel for manylinux_2_28_aarch64	2m 8s
Build CPU wheel for manylinux_2_28_x86_64	2m 24s
Build CPU wheel for manylinux2014_aarch64	1m 56s
Run Python tests (cpu-arm64)	1m 23s
Run Python tests (cpu-amd64)	20m 38s
Run Python tests (multiple-gpu)	20m 41s
Run Python tests (single-gpu)	21m 22s
Run Google Tests with GPUs (Suite gpu-rmm, Runner linux-amd64-gpu)	5m 18s
Run Google Tests with GPUs (Suite mgpu, Runner linux-amd64-mgpu)	3m 0s
Run Google Tests with GPUs (Suite gpu, Runner linux-amd64-gpu)	5m 6s
	2h 45m 26s

Provide feedback