Uncouple gridspec and layout and use native `set_layout_engine` present in MPL

[cvanelteren]

The layout mechanism is currently buggy for irregular grids (#306). In order to improve readability, I propose using a new LayoutManager that we can use an set using the appropriate layout methods present in Figure.

[cvanelteren]

I've investigated this and implementing it for UltraPlot would be quite challenging. Here's why:

The Core Problem

In ProPlot, the GridSpec class is fundamental to how layouts work—it effectively flattens the layout structure. However, this approach makes us incompatible with matplotlib because we've overridden gridspec in ways that differ substantially from the standard implementation.

Layout Engines in Matplotlib

Matplotlib offers several layout engines to automatically adjust subplot spacing and prevent overlapping elements:

• tight_layout - Basic automatic spacing adjustment
• constrained_layout - More sophisticated solver that handles complex subplot arrangements, colorbars, and legends

These engines work with matplotlib's standard gridspec structure, which is why we can't currently use them.

The Dilemma

I'm uncertain whether it's worth reverting our custom gridspec implementation just to add our layout algorithm as an alternative to matplotlib's layout engines. Retooling the GridSpec to be compatible could actually solve several existing issues (#188, #270, and #297).

Why This Matters

Issue #297 is particularly important—I've encountered it many times myself. It essentially requires us to provide a constrained_layout solver, which matplotlib already has. But since our gridspec isn't compatible with matplotlib's, we can't leverage their layout engines. This is a significant limitation in my opinion.

[cvanelteren]

pinging @beckermr for visibility^