Improve the GC harness Summary output

[eightbitraptor]

This PR adds more information about GC counts and timing to the summary output of the GC handler to avoid having to scroll up to the individual tables and infer information. It adds:

• Ratios of the average total time spent marking/sweeping. So the ratio is mean(total_control_time) / mean(total_experiment_time). Above 1 represents a speed-up in the experiment branch.
• Ratios of the average time spend marking/sweeping per GC (across all GC's major and minor).. Each branch is calculated using mean(total_marking_time) / (mean(major_count) + mean(minor_count))
• Comparison values of average major and minor count per branch. 1.4 -> 6.0 says that the control branch ran an average of 1.4 GC cycles across all runs, and the experiment ran an average of 6.0.
• Percentage of total GC time spent doing Minor GC's per branch. Intended to give an indication as to what the major/minor time split per branch is.

the time / iter and time / GC ratios are intended to help determine whether the marking/sweeping time improved because marking/sweeping got faster, or because we're running fewer. For example, If "mark time/iter" improves but "mark time/GC" stays flat then you know that the win came from running fewer GC cycles.

New Summary output looks like this:

master: ruby 4.1.0dev (2026-06-17T10:00:21Z master 45fd5d58f1) +YJIT +PRISM [x86_64-linux]
experiment: ruby 4.1.0dev (2026-06-17T10:00:21Z mvh-obspace-malloc.. 45fd5d58f1) +PRISM [x86_64-linux]

----------  -------------  ------------  ---------------  ------------  ------------------  -----------------  ---------------------
bench         master (ms)     RSS (MiB)  experiment (ms)     RSS (MiB)  experiment 1st itr  master/experiment  RSS master/experiment
lobsters     977.1 ± 1.2%  352.3 ± 0.0%    1760.3 ± 5.0%  307.2 ± 0.2%               1.588              0.555                  1.147
railsbench  1838.6 ± 1.5%  139.1 ± 0.3%    3506.9 ± 0.5%  121.1 ± 1.2%               0.728              0.524                  1.149
gcbench      699.4 ± 5.8%   47.3 ± 0.0%    2904.4 ± 1.4%   46.8 ± 0.0%               0.253              0.241                  1.010
----------  -------------  ------------  ---------------  ------------  ------------------  -----------------  ---------------------

GC summary:
----------  --------------  ---------------  ------------  -------------  -------------  -------------  -------------
bench       mark time/iter  sweep time/iter  mark time/GC  sweep time/GC     major/iter     minor/iter     minor GC %
lobsters             0.957            0.929         0.957          0.929   0.0  →   0.0   0.9  →   0.9  100%  →  100%
railsbench           0.179            1.020         0.134          0.768   0.2  →   1.8   9.5  →   5.5   98%  →   75%
gcbench              0.963            0.987         0.963          0.987   0.0  →   0.0  52.0  →  52.0  100%  →  100%
----------  --------------  ---------------  ------------  -------------  -------------  -------------  -------------

Legend:
- experiment 1st itr: ratio of master/experiment time for the first benchmarking iteration.
- master/experiment: ratio of master/experiment time. Higher is better for experiment. Above 1 represents a speedup.
- RSS master/experiment: ratio of master/experiment RSS. Higher is better for experiment. Above 1 means lower memory usage.
- GC summary compares master → comparison. Ratio columns are master/comparison; above 1 means the comparison spent less GC time.
- mark time/iter and sweep time/iter compare total GC phase time per benchmark iteration, so they include both per-GC cost and GC frequency changes.
- mark time/GC and sweep time/GC compare average phase time per GC, isolating whether each GC became cheaper or more expensive.
- major/iter, minor/iter, and minor GC % show master → comparison values, not ratios. Rows with no GC activity are omitted.

[peterzhu2118]

Nice, I really like this!