✨ Lazy generation of chart data, 'Artist' registry, and better SVG export

Now all chart-data parameters, including the overlayed zones, are included in the ChartConfig,
and curve-generation only happens when chart is plotted (or saved to disk, or if chart.process_chart() is invoked 😜).

Any change in the chart configuration will trigger a data-regen for all items before plotting.

When plotting over some matplotlib Axes, all objects created now have meaningful (and deterministic) ids,
and are tracked and accessible under chart.artists, for easier customization of single matplotlib objects,
also adding a recognizable hierarchy in generated SVGs, (enabling potential for CSS over SVG charts 🌈)

Changes

• ✨ Name each matplotlib Artist obj in plot with custom readable IDs, for easier recognition and readable object hierarchy in SVGs, and maintain a registry of all objects in plot to access them by kind under chart.artists property, allowing fine-grain control to customize all details in the psychrochart
• ♻️ Refactor chart 'zones' and store zone definitions in ChartConfig, to regenerate plot patches (closed PsychroCurve, with ZoneStyle) on-demand, same as the other chart curves, generating new 'xy' points when config is changed. If ChartParams.with_zones is enabled, but there are no zones in config when chart is created, the default winter/summer 'dbt-rh' zones are added.
• ⚡️ Add lazy generation of psychro curves to avoid updating the chart data until it's needed to plot, and add a chart.process_chart() method to trigger data regeneration if the chart config has changed or there is no chart data yet
• 💥 Make chart.saturation a single PsychroCurve, instead of a collection of curves
• ✅ tests: Add unit tests for new ChartRegistry and adapt to behaviour changes and readable hierarchy in SVG output
• 📝 Update example notebook showing chart.artists usage
• 🚀 env: Update deps, bump version, and update CHANGELOG