[Feature] port configs from upstream

[Locietta]

Porting kernel config from upstream. Hope to resolve the performance regression (#8).

Various kernel feature enabled or modified:

• BBRv2 + FQ_PIE for networking
• some performance feature to SLAB
• Multi-Gen LRU (better performance with limited mem)
• UCLAMP and auto task group
• NUMA
• preemption model from none to voluntary
• ZSTD kernel compression

[Locietta]

Seems like new configs mostly fix the performance regression on process creation and syscall overhead.

Voluntary preemption and UCLAMP task group makes huge improvement in multi-core process creation. Though don't know how syscall overhead is reduced, but it somehow works. 🤔

ZSTD compression makes kernel image much smaller btw.

raw benchmark result: new_benchmark.txt