feat(nvme): add NVMe driver support

[valopok]

I added NVMe driver support by using an adapted version of the vroom driver.
Alongside this pull request, I will open another pull request in hermit-rs to add an example testing the driver.

[github-actions]

Benchmark Results

Benchmark	Current: e18e728	Previous: 0b7fc73	Performance Ratio
startup_benchmark Build Time	110.38 s	124.30 s	0.89
startup_benchmark File Size	0.86 MB	0.86 MB	1.00
Startup Time - 1 core	0.88 s (±0.03 s)	1.05 s (±0.02 s)	0.84
Startup Time - 2 cores	0.89 s (±0.03 s)	1.04 s (±0.03 s)	0.86
Startup Time - 4 cores	0.92 s (±0.04 s)	1.03 s (±0.02 s)	0.89
multithreaded_benchmark Build Time	112.48 s	126.23 s	0.89
multithreaded_benchmark File Size	0.96 MB	0.96 MB	1.00
Multithreaded Pi Efficiency - 2 Threads	88.15 % (±9.45 %)	84.80 % (±7.75 %)	1.04
Multithreaded Pi Efficiency - 4 Threads	43.40 % (±3.60 %)	42.45 % (±2.98 %)	1.02
Multithreaded Pi Efficiency - 8 Threads	25.29 % (±1.61 %)	24.52 % (±2.36 %)	1.03
micro_benchmarks Build Time	269.90 s	274.29 s	0.98
micro_benchmarks File Size	0.97 MB	0.97 MB	1.00
Scheduling time - 1 thread	151.59 ticks (±22.22 ticks)	148.36 ticks (±12.93 ticks)	1.02
Scheduling time - 2 threads	98.97 ticks (±16.66 ticks)	88.04 ticks (±13.54 ticks)	1.12
Micro - Time for syscall (getpid)	9.00 ticks (±4.20 ticks)	10.38 ticks (±5.26 ticks)	0.87
Memcpy speed - (built_in) block size 4096	54207.00 MByte/s (±40102.58 MByte/s)	56952.43 MByte/s (±41782.00 MByte/s)	0.95
Memcpy speed - (built_in) block size 1048576	14882.60 MByte/s (±12012.19 MByte/s)	15647.95 MByte/s (±12805.19 MByte/s)	0.95
Memcpy speed - (built_in) block size 16777216	11040.15 MByte/s (±8996.52 MByte/s)	11222.99 MByte/s (±9160.99 MByte/s)	0.98
Memset speed - (built_in) block size 4096	54611.11 MByte/s (±40392.92 MByte/s)	57346.55 MByte/s (±42024.24 MByte/s)	0.95
Memset speed - (built_in) block size 1048576	15266.09 MByte/s (±12240.59 MByte/s)	16072.80 MByte/s (±13089.20 MByte/s)	0.95
Memset speed - (built_in) block size 16777216	11216.34 MByte/s (±9076.36 MByte/s)	11485.66 MByte/s (±9311.28 MByte/s)	0.98
Memcpy speed - (rust) block size 4096	53535.93 MByte/s (±39338.23 MByte/s)	56725.21 MByte/s (±40857.65 MByte/s)	0.94
Memcpy speed - (rust) block size 1048576	15169.73 MByte/s (±12220.52 MByte/s)	15912.94 MByte/s (±12954.57 MByte/s)	0.95
Memcpy speed - (rust) block size 16777216	11109.98 MByte/s (±9047.33 MByte/s)	11044.74 MByte/s (±8978.39 MByte/s)	1.01
Memset speed - (rust) block size 4096	54501.80 MByte/s (±39992.32 MByte/s)	57139.62 MByte/s (±41108.43 MByte/s)	0.95
Memset speed - (rust) block size 1048576	15444.76 MByte/s (±12366.77 MByte/s)	16296.05 MByte/s (±13185.87 MByte/s)	0.95
Memset speed - (rust) block size 16777216	11293.89 MByte/s (±9132.20 MByte/s)	11283.59 MByte/s (±9110.64 MByte/s)	1.00
alloc_benchmarks Build Time	265.29 s	276.44 s	0.96
alloc_benchmarks File Size	0.92 MB	0.92 MB	1.00
Allocations - Allocation success	100.00 %	100.00 %	1
Allocations - Deallocation success	100.00 %	100.00 %	1
Allocations - Pre-fail Allocations	100.00 %	100.00 %	1
Allocations - Average Allocation time	17277.93 Ticks (±618.42 Ticks)	17424.87 Ticks (±622.62 Ticks)	0.99
Allocations - Average Allocation time (no fail)	17277.93 Ticks (±618.42 Ticks)	17424.87 Ticks (±622.62 Ticks)	0.99
Allocations - Average Deallocation time	2156.44 Ticks (±446.74 Ticks)	2128.52 Ticks (±372.69 Ticks)	1.01
mutex_benchmark Build Time	267.07 s	281.32 s	0.95
mutex_benchmark File Size	0.97 MB	0.97 MB	1.00
Mutex Stress Test Average Time per Iteration - 1 Threads	32.48 ns (±3.06 ns)	33.54 ns (±3.06 ns)	0.97
Mutex Stress Test Average Time per Iteration - 2 Threads	27.10 ns (±2.96 ns)	28.50 ns (±2.47 ns)	0.95

This comment was automatically generated by workflow using github-action-benchmark.

[jounathaen]

Hi @valopok and thank you for the contribution.
This is an interesting idea. Yet, I'm a little unsure, if this fit's the goals of Hermit, as I personally also see a value in keeping the codebase as small as possible. So, could you maybe give us a little motivation for the addition of this driver?

I'd also be curious, if it could alternatively be implemented outside the kernel and which interfaces to the kernel are missing if it isn't possible at the moment.