Include run label in directory name

[frazane]

This pull request updates how run identifiers and output paths are constructed for run data and output results.

Summary of changes

• the run output directory name is given a more human-readable name e.g. forecaster-d084-91a251cd, interpolator-8d1e-on-forecaster-d084-1fa265c9, interpolator-8d1e-on-ana-1fa265c9
• simplify structure of results output directories
• reduce verbosity of snakemake console output, while improving console output form evalml
• move "expensive" outputs to data dir for caching purposes

[jonasbhend]

We should find a better solution than to trigger a complete rerun of everything if the label of a forecast run or baseline changes. This will come back to bite us when iterating to produce production-ready plots for papers, presentations etc.

I totally see the need to render plots in showcases and experiments more easily identifiable and think using the label there is probably the low-hanging fruit, but can we please avoid doing this for the inference and verification in output/data/...

[dnerini]

new proposition is to distinguish more clearly between

• output/data is a cache --> not human readble
• output/results is main interface to user --> human readable

meaning that there is no need for caching results. Each evalml run creates a new results folde ordered by timestamp and if something needs caching because it's expensive (eg plotting all frames for showcases) it can be moved to output/data.

[dnerini]

PR 's ready for review @MeteoSwiss/s-ruc

example output:

and in the console:

(evalml) ned@balfrin-ln001:~/src/evalmldev> evalml showcase config/interpolators-co2.yaml
host: balfrin-ln001
Building DAG of jobs...
SLURM run ID: 16c2ff94-9282-4ef9-bbaf-15266c9c316c
Using shell: /usr/local/bin/bash
Provided remote nodes: 50
Provided resources: gpus=16

Job stats:
job                               count
------------------------------  -------
create_inference_pyproject            3
create_inference_venv                 3
execute_inference                    12
make_forecast_animation              72
make_squashfs_image                   3
plot_forecast_frame                6312
plot_meteogram                       72
prepare_inference_forecaster          4
prepare_inference_interpolator        8
showcase_all                          1
write_summary                         3
total                              6493
────────────────────────────────────────────────────────────────────────
🚀 EvalML workflow started
   Time:        2026-01-29 14:07:12
   Snakemake:   snakemake --executor slurm --resources gpus=16 --default-resources slurm_partition=postproc cpus_per_task=1 mem_mb_per_cpu=1800 runtime=1h --jobs 50 --groups plot_forecast_frame=plot_forecast_frame --group-components plot_forecast_frame=32 --configfile config/interpolators-co2.yaml --cores 4 showcase_all --quiet rules

   Workdir:     /users/ned/src/evalmldev
   Config:      interpolators-co2.yaml
   Experiment:  20260129_interpolators-co2_8719abb0
   Candidates:  interpolator-8d1e-on-analysis-a712b892, forecaster-d084-91a251cd, interpolator-8d1e-on-forecaster-d084-1fa265c9
   Data dir:    output/data
   Logs dir:    output/logs
   Results dir: output/results/20260129_interpolators-co2_8719abb0
────────────────────────────────────────────────────────────────────────

Select jobs to execute...
Execute 3 jobs...
[...]
373 of 374 steps (99.7%) done
Select jobs to execute...
Execute 1 jobs...
[Thu Jan 29 13:54:54 2026]
Finished jobid: 0 (Rule: showcase_all)
374 of 374 steps (100%) done
Cleaning up log files older than 10 day(s).

Complete log(s): /users/ned/src/evalmldev/.snakemake/log/2026-01-29T133007.301999.snakemake.log
────────────────────────────────────────────────────────────────────────
✅ Workflow finished successfully
   Duration:  25m 17s
   Results:   output/results/20260129_interpolators-co2_9afe91fb
────────────────────────────────────────────────────────────────────────

[jonasbhend]

I won't have time to do a proper review, but I like the proposed changes.