refactor(fs): refactor path traversal handling

[fogti]

I realized that by "just" factoring out all path traversal handling, I don't even need to introduce an std::path::Path equivalent.

The afaik only disadvantage that this has is the re-allocations during traversal into fuse and uhyve filesystems.

[github-actions]

Benchmark Results

Details

Benchmark	Current: 6c0b8d6	Previous: e4ac074	Performance Ratio
startup_benchmark Build Time	98.97 s	98.80 s	1.00 ❗
startup_benchmark File Size	0.86 MB	0.82 MB	1.06 ❗
Startup Time - 1 core	0.95 s (±0.04 s)	0.94 s (±0.03 s)	1.00
Startup Time - 2 cores	0.94 s (±0.03 s)	0.96 s (±0.03 s)	0.98
Startup Time - 4 cores	0.98 s (±0.03 s)	0.94 s (±0.03 s)	1.03
multithreaded_benchmark Build Time	101.60 s	100.58 s	1.01 ❗
multithreaded_benchmark File Size	0.91 MB	0.97 MB	0.94 ❗
Multithreaded Pi Efficiency - 2 Threads	85.94 % (±8.07 %)	90.29 % (±8.49 %)	0.95
Multithreaded Pi Efficiency - 4 Threads	42.10 % (±3.63 %)	43.48 % (±3.34 %)	0.97
Multithreaded Pi Efficiency - 8 Threads	24.81 % (±1.47 %)	25.20 % (±1.85 %)	0.98
micro_benchmarks Build Time	93.89 s	163.78 s	0.57 ❗
micro_benchmarks File Size	0.91 MB	0.97 MB	0.94 ❗
Scheduling time - 1 thread	62.67 ticks (±2.95 ticks)	99.58 ticks (±31.86 ticks)	0.63 ❗
Scheduling time - 2 threads	35.22 ticks (±5.09 ticks)	55.55 ticks (±16.94 ticks)	0.63 ❗
Micro - Time for syscall (getpid)	4.12 ticks (±0.60 ticks)	7.42 ticks (±3.84 ticks)	0.56
Memcpy speed - (built_in) block size 4096	68722.93 MByte/s (±49083.95 MByte/s)	60347.83 MByte/s (±42940.07 MByte/s)	1.14
Memcpy speed - (built_in) block size 1048576	29656.47 MByte/s (±24257.13 MByte/s)	25535.97 MByte/s (±21900.21 MByte/s)	1.16
Memcpy speed - (built_in) block size 16777216	28612.11 MByte/s (±23818.23 MByte/s)	21177.04 MByte/s (±17856.22 MByte/s)	1.35
Memset speed - (built_in) block size 4096	69305.46 MByte/s (±49392.76 MByte/s)	60998.96 MByte/s (±43392.21 MByte/s)	1.14
Memset speed - (built_in) block size 1048576	30422.07 MByte/s (±24690.11 MByte/s)	26412.57 MByte/s (±22471.00 MByte/s)	1.15
Memset speed - (built_in) block size 16777216	29382.80 MByte/s (±24253.80 MByte/s)	21301.98 MByte/s (±17886.43 MByte/s)	1.38
Memcpy speed - (rust) block size 4096	61039.48 MByte/s (±45117.97 MByte/s)	53768.42 MByte/s (±39033.99 MByte/s)	1.14
Memcpy speed - (rust) block size 1048576	29457.03 MByte/s (±24212.74 MByte/s)	24336.19 MByte/s (±21041.44 MByte/s)	1.21
Memcpy speed - (rust) block size 16777216	28579.90 MByte/s (±23776.20 MByte/s)	20742.62 MByte/s (±17390.56 MByte/s)	1.38
Memset speed - (rust) block size 4096	61828.95 MByte/s (±45590.53 MByte/s)	54278.77 MByte/s (±39313.56 MByte/s)	1.14
Memset speed - (rust) block size 1048576	30234.34 MByte/s (±24655.96 MByte/s)	25118.65 MByte/s (±21537.32 MByte/s)	1.20
Memset speed - (rust) block size 16777216	29351.50 MByte/s (±24215.94 MByte/s)	20922.60 MByte/s (±17443.37 MByte/s)	1.40
alloc_benchmarks Build Time	91.89 s	150.20 s	0.61 ❗
alloc_benchmarks File Size	0.94 MB	0.89 MB	1.05 ❗
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	5848.19 Ticks (±229.18 Ticks)	10243.26 Ticks (±359.99 Ticks)	0.57 ❗
Allocations - Average Allocation time (no fail)	5848.19 Ticks (±229.18 Ticks)	10243.26 Ticks (±359.99 Ticks)	0.57 ❗
Allocations - Average Deallocation time	751.65 Ticks (±111.38 Ticks)	2287.98 Ticks (±940.40 Ticks)	0.33 ❗
mutex_benchmark Build Time	91.88 s	152.42 s	0.60 ❗
mutex_benchmark File Size	0.91 MB	0.97 MB	0.94 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	12.56 ns (±0.70 ns)	18.06 ns (±4.09 ns)	0.70 ❗
Mutex Stress Test Average Time per Iteration - 2 Threads	13.06 ns (±0.79 ns)	19.90 ns (±2.71 ns)	0.66 ❗

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

[fogti]

It might be a good idea to instead only store the part in prefix that is actually "beyond" original_prefix.

[fogti]

I had to change this method from a recursion to a loop to avoid having to allocate a Box for the .stat (because otherwise calling .await on it inside of the function would lead to an infinitely-sized closure).

[fogti]

TODO: introduce traverse_multiple.