Mh/diffusion full pipeline latent rmse and diagnostics

[moritzhauschulz]

Description

Implements plotting of latent spectra and decoded intermediate samples during inference. Diagnostics tool only. Should be merged on top of #2683.

# --- Per-ODE-step map & spectrum diagnostics (inference, single-step forecasting only) ---
# Decodes x_t and x0_hat at each ODE step and plots maps (x_t | x0_hat | decode(z) | truth) plus
# angular power spectra in latent and physical space, under <results>/<run_id>/plots/ode_diagnostics.
# Requires training_config.forecast.num_steps=1 (rollout leaves no reference target). Off by default.
diag_ode_maps: false           # master switch; costs 2 decoder passes + 4 map panels per step
diag_ode_every_n_steps: 1      # 1 = every ODE step; raise to cut render time
diag_channels: ["2t", "q_850"]
diag_stream: ERA5
diag_latent_channels: 128      # latent channels sampled for the mean C_l (0 = all)

Issue Number

Checklist before asking for review

• I have performed a self-review of my code
• My changes comply with basic sanity checks:
  • I have fixed formatting issues with ./scripts/actions.sh lint
  • I have run unit tests with ./scripts/actions.sh unit-test
  • I have documented my code and I have updated the docstrings.
  • I have added unit tests, if relevant
• I have tried my changes with data and code:
  • I have run the integration tests with ./scripts/actions.sh integration-test
  • (bigger changes) I have run a full training and I have written in the comment the run_id(s): launch-slurm.py --time 60
  • (bigger changes and experiments) I have shared a hegdedoc in the github issue with all the configurations and runs for this experiments
• I have informed and aligned with people impacted by my change:
  • for config changes: the MatterMost channels and/or a design doc
  • for changes of dependencies: the MatterMost software development channel