Convert dev mode unit tests to integration tests

[flaub]

This avoids env var concurrency issues in unit tests

[github-actions]

Benchmark for Linux-cuda c0ab4cb

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	5.1±0.10ms	5.1±0.10ms	0.00%
fib/100/prove	777.2±2.62ms	740.8±0.87ms	-4.68%
fib/100/total	787.8±3.99ms	747.0±3.74ms	-5.18%
fib/1000/execute	5.6±0.11ms	5.5±0.10ms	-1.79%
fib/1000/prove	805.4±3.03ms	772.8±1.68ms	-4.05%
fib/1000/total	815.2±4.03ms	775.0±2.53ms	-4.93%
fib/10000/execute	10.3±0.16ms	10.1±0.11ms	-1.94%
fib/10000/prove	3.4±0.01s	2.8±0.01s	-17.65%
fib/10000/total	3.4±0.01s	2.8±0.01s	-17.65%

Benchmark for Linux-default c0ab4cb

Click to hide benchmark

Test	Base	PR	%
fib/100/execute	4.8±0.19ms	4.8±0.14ms	0.00%
fib/100/prove	2.6±0.35s	1693.4±38.60ms	-34.87%
fib/100/total	1738.5±121.83ms	1701.4±5.92ms	-2.13%
fib/1000/execute	5.3±0.12ms	5.2±0.11ms	-1.89%
fib/1000/prove	2.4±0.36s	1716.7±7.47ms	-28.47%
fib/1000/total	1746.8±80.66ms	1745.9±57.88ms	-0.05%
fib/10000/execute	9.1±0.18ms	9.0±0.10ms	-1.10%
fib/10000/prove	7.1±0.60s	6.4±0.22s	-9.86%
fib/10000/total	6.5±0.21s	6.4±0.22s	-1.54%

Benchmark for macOS-default

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

Benchmark for macOS-metal

    <details open>
      <summary>Click to hide benchmark</summary>
      Benchmarks have changed between the two branches, unable to diff.
    </details>

[SchmErik]

thanks for doing this!