running analysis from gui does not create config.used.yaml and run_info.json

[Dahaniel]

running analysis from gui does not create config.used.yaml and run_info.json

[degoldschmidt]

Reproduced and fixed in #4.

The cause was duplication rather than a broken function: flypad run and the GUI each had
their own copy of the pipeline sequence, and only the CLI copy went on to write the
provenance pair. Both now drive a single pipeline.run_experiment(), so the results
directory is the same whichever produced it — run_info.json records which, in its
command field (run or gui).

One knock-on worth knowing about

flypad stats <results_dir> reuses the run's config.used.yaml when it is there and
falls back to a default config when it is not, then overwrites per_condition and
comparisons in that directory. So running flypad stats on a GUI results directory
silently used default settings instead of yours.

In practice this was narrower than it sounds: the default config agrees with both presets
on every field stats actually reads (stats.ci_level, n_permutations, statistic and
plotting.facet_by), so a run with a stock config recomputed to the same numbers. It only
diverged if your experiment.yaml overrode one of those. If it did, and you ran
flypad stats over a GUI directory, that directory's per_condition.csv /
comparisons.csv were regenerated with defaults — re-running flypad stats after this
lands will pick up your settings and correct them.

Checking it

uv run flypad gui   # run an analysis as usual, then:
ls results/run_info.json results/config.used.yaml
python -c "import json; print(json.load(open('results/run_info.json'))['command'])"   # -> gui

Thanks for reporting both this and #2.