Fixes risc0-build deadlocks on large outputs

[mothran]

Previously if we had a large volume of cargo build errors it would hit the piped limit of the child process terminate. By calling cmd.status() we would invoke a second version of the subprocess. Using the spawn() return value and .wait() means we correctly pull from the right piped IO.

This might have also fixed double-invoking the cargo build child process but I am not 100% sure on that.

[github-actions]

Benchmark for Linux-cuda 75910c6

Click to hide benchmark

Test	Base	PR	%
fib/100/proof	855.2±40.81ms	820.6±45.43ms	-4.05%
fib/100/run	337.6±9.64ms	328.1±10.15ms	-2.81%
fib/200/proof	850.2±50.34ms	800.4±38.64ms	-5.86%
fib/200/run	332.2±7.86ms	325.7±6.96ms	-1.96%

Benchmark for Linux-default 75910c6

Click to hide benchmark

Test	Base	PR	%
fib/100/proof	2.3±0.02s	2.3±0.02s	0.00%
fib/100/run	344.6±10.14ms	331.2±0.90ms	-3.89%
fib/200/proof	2.3±0.02s	2.3±0.02s	0.00%
fib/200/run	351.8±19.91ms	334.9±2.50ms	-4.80%

Benchmark for macOS-default 75910c6

Click to hide benchmark

Test	Base	PR	%
fib/100/proof	1593.4±9.41ms	1593.2±9.39ms	-0.01%
fib/100/run	110.3±0.59ms	110.3±0.35ms	0.00%
fib/200/proof	1609.2±29.00ms	1602.2±15.23ms	-0.43%
fib/200/run	112.1±0.61ms	112.0±0.44ms	-0.09%

Benchmark for macOS-metal 75910c6

Click to hide benchmark

Test	Base	PR	%
fib/100/proof	464.9±4.46ms	461.5±3.17ms	-0.73%
fib/100/run	113.7±0.45ms	111.9±0.89ms	-1.58%
fib/200/proof	467.2±4.03ms	465.4±3.95ms	-0.39%
fib/200/run	115.4±0.33ms	113.9±0.55ms	-1.30%