test: measure VC discharging separately in Sym mvcgen benchmarks (#12551)
This PR refactors the benchmark driver of the Sym mvcgen benchmarks such
that time spent for discharging VCs and instantiation of MVars is
measured separately from VC generation.
Example output:
```
baseline_add_sub_cancel
goal_100: 57 ms, 0 VCs, kernel: 22 ms
goal_500: 353 ms, 0 VCs, kernel: 160 ms
goal_1000: 755 ms, 0 VCs, kernel: 437 ms
vcgen_add_sub_cancel
goal_100: 36 ms, 1 VCs by grind: 21 ms, kernel: 35 ms
goal_500: 149 ms, 1 VCs by grind: 115 ms, kernel: 214 ms
goal_1000: 314 ms, 1 VCs by grind: 249 ms, kernel: 478 ms
vcgen_deep_add_sub_cancel
goal_100: 65 ms, 1 VCs by grind: 23 ms, kernel: 82 ms
goal_500: 262 ms, 1 VCs by grind: 123 ms, kernel: 539 ms
goal_1000: 611 ms, 1 VCs by grind: 292 ms, kernel: 1075 ms
vcgen_get_throw_set
goal_100: 87 ms, 101 VCs by sorry: 16 ms, kernel: 93 ms
goal_500: 332 ms, 501 VCs by sorry: 289 ms, instantiate > 1000ms: 23363 ms, kernel: 770 ms
goal_1000: 794 ms, 1001 VCs by sorry: 1332 ms, instantiate > 1000ms: 334614 ms, kernel: 1882 ms
```
c0b9ff1148