feat: add uhyve feature

[mkroening]

Currently, Hermit unikernels can only be specialized to which device drivers to include, but not to which VMM the unikernel is run in.

This PR enables such specialization by feature-gating the Uhyve code. This means when Uhyve is targeted, the uhyve feature must be enabled. Currently, this only reduces code size and dependencies for unikernels that do not target Uhyve. In a follow-up PR, I will add the loader feature to enable savings for unikernels that do target Uhyve.

In the grander scheme of things, going this direction allows us to nicely specialize for direct PVH support in the future (#2366).

Depends on hermit-os/hermit-rs#976.
I will pull out some of the preparatory refactorings into a separate PR.

[github-actions]

Benchmark Results

Details

Benchmark	Current: 6bfbe18	Previous: e0b9d04	Performance Ratio
startup_benchmark Build Time	89.57 s	90.22 s	0.99 ❗
startup_benchmark File Size	0.76 MB	0.80 MB	0.95 ❗
Startup Time - 1 core	0.81 s (±0.03 s)	0.82 s (±0.04 s)	0.98
Startup Time - 2 cores	0.79 s (±0.03 s)	0.82 s (±0.04 s)	0.96
Startup Time - 4 cores	0.80 s (±0.03 s)	0.83 s (±0.03 s)	0.97
multithreaded_benchmark Build Time	90.60 s	92.03 s	0.98 ❗
multithreaded_benchmark File Size	0.86 MB	0.91 MB	0.95 ❗
Multithreaded Pi Efficiency - 2 Threads	88.96 % (±7.32 %)	89.53 % (±8.21 %)	0.99
Multithreaded Pi Efficiency - 4 Threads	44.73 % (±2.78 %)	43.81 % (±2.78 %)	1.02
Multithreaded Pi Efficiency - 8 Threads	25.91 % (±1.71 %)	26.30 % (±1.72 %)	0.99
micro_benchmarks Build Time	100.39 s	100.06 s	1.00 ❗
micro_benchmarks File Size	0.87 MB	0.92 MB	0.95 ❗
Scheduling time - 1 thread	69.47 ticks (±4.33 ticks)	73.83 ticks (±4.75 ticks)	0.94
Scheduling time - 2 threads	39.99 ticks (±5.33 ticks)	40.31 ticks (±4.87 ticks)	0.99
Micro - Time for syscall (getpid)	3.66 ticks (±0.33 ticks)	3.03 ticks (±0.28 ticks)	1.21 ❗
Memcpy speed - (built_in) block size 4096	76314.72 MByte/s (±52876.24 MByte/s)	74297.66 MByte/s (±51450.89 MByte/s)	1.03
Memcpy speed - (built_in) block size 1048576	30314.72 MByte/s (±24918.65 MByte/s)	30384.44 MByte/s (±25123.41 MByte/s)	1.00
Memcpy speed - (built_in) block size 16777216	25221.33 MByte/s (±21017.29 MByte/s)	22358.71 MByte/s (±18609.52 MByte/s)	1.13
Memset speed - (built_in) block size 4096	76715.67 MByte/s (±53169.65 MByte/s)	74713.95 MByte/s (±51707.68 MByte/s)	1.03
Memset speed - (built_in) block size 1048576	31076.29 MByte/s (±25340.73 MByte/s)	31139.74 MByte/s (±25543.05 MByte/s)	1.00
Memset speed - (built_in) block size 16777216	25971.48 MByte/s (±21497.68 MByte/s)	23007.68 MByte/s (±19039.40 MByte/s)	1.13
Memcpy speed - (rust) block size 4096	68790.94 MByte/s (±47972.03 MByte/s)	66767.89 MByte/s (±46628.30 MByte/s)	1.03
Memcpy speed - (rust) block size 1048576	30322.52 MByte/s (±24811.22 MByte/s)	30165.31 MByte/s (±25004.23 MByte/s)	1.01
Memcpy speed - (rust) block size 16777216	25638.70 MByte/s (±21348.28 MByte/s)	23346.06 MByte/s (±19388.66 MByte/s)	1.10
Memset speed - (rust) block size 4096	68974.54 MByte/s (±48103.60 MByte/s)	67218.70 MByte/s (±46992.71 MByte/s)	1.03
Memset speed - (rust) block size 1048576	31092.23 MByte/s (±25244.77 MByte/s)	30928.81 MByte/s (±25426.89 MByte/s)	1.01
Memset speed - (rust) block size 16777216	26392.92 MByte/s (±21824.40 MByte/s)	24055.06 MByte/s (±19863.24 MByte/s)	1.10
alloc_benchmarks Build Time	91.32 s	96.62 s	0.95 ❗
alloc_benchmarks File Size	0.83 MB	0.87 MB	0.96 ❗
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	7047.86 Ticks (±130.61 Ticks)	6884.16 Ticks (±146.01 Ticks)	1.02
Allocations - Average Allocation time (no fail)	7047.86 Ticks (±130.61 Ticks)	6884.16 Ticks (±146.01 Ticks)	1.02
Allocations - Average Deallocation time	1529.14 Ticks (±629.17 Ticks)	1212.05 Ticks (±597.54 Ticks)	1.26
mutex_benchmark Build Time	92.09 s	94.75 s	0.97 ❗
mutex_benchmark File Size	0.87 MB	0.92 MB	0.95 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	13.02 ns (±0.88 ns)	13.58 ns (±0.83 ns)	0.96
Mutex Stress Test Average Time per Iteration - 2 Threads	16.74 ns (±12.84 ns)	16.88 ns (±6.68 ns)	0.99

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

[jounathaen]

Can we make uhyve a default feature?