Add residual plotting

[choutkaj]

Summary

This PR adds residual plotting to the Axes-first plotting API.

Included:

• Adds standalone plot_residuals()
• Computes residuals as observed - predicted
• Supports aggregated residuals, matching the default technical-replicate aggregation used in plotting observations
• Supports raw replicate residuals with aggregate=False
• Supports compound and experiment filtering
• Adds optional zero reference line, enabled by default
• Supports log or linear x-axis scaling
• Labels residual units using DoseResponseData.response_unit
• Exports plot_residuals() from bindcurve.plotting and top-level bindcurve
• Adds tests for:
  • aggregated residual plotting
  • raw replicate residual plotting
  • disabling the zero line
  • linear x-axis scaling
  • missing successful fits

Notes

This PR keeps residual plotting as a standalone helper so users can compose diagnostics flexibly:

fig, ax = plt.subplots()
bc.plot_residuals(data, results, ax=ax)

It does not add a multi-panel residual layout yet. That can come later as a compound/grid plotting layer.

Test plan

• Existing pytest workflow should run the expanded plotting tests
• Existing Ruff workflow should lint the plotting package and tests
• Branch is cleanly ahead of main with behind_by=0

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dbbf3d5987

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Handle compound-summary fits when filtering residual rows

plot_residuals() drops all points for fits produced with FitSettings(strategy="per_compound_summary") because those fits have experiment_id="compound_summary", but raw observation tables only contain real experiment IDs (e.g., exp1, exp2). This filter makes fit_table empty, so the loop silently skips plotting residuals and users get an empty diagnostic despite having a successful fit. Please special-case summary fits (or any fit ID not present in the table) so residuals are computed from the aggregated compound-level observations instead of filtering everything out.

Useful? React with 👍 / 👎.