Jupyter-scatter

- Fix: update `color`, `opacity`, and `size` scales as the domains update
- Fix: auto-reset `x` and `y` scale domains upon updating the `x` and `y` data
- Fix: use better number formatter for the legend

- Add the ability to show a tooltip upon hovering over a point via `scatter.tooltip(true)` ([86](https://github.com/flekschas/jupyter-scatter/pull/86))
- Fix axes updating of linked scatter plots when panning and zooming ([87](https://github.com/flekschas/jupyter-scatter/issues/87))
- Fix missing x-axes of linked scatter plots ([84](https://github.com/flekschas/jupyter-scatter/issues/84))
- Fix a type in the return value of `scatter.xy()`

- Fix: Prevent resetting the filter upon color, size, or opacity changes
- Fix: Upon changing the associated data frame via `scatter.data(new_df)`, reapply color, size, and opacity settings

- Add ability to specify titles when composing multiple scatter plots by passing tuples of `(Scatter, title)` to `compose()` or `link()`
- Fix: Add docstrings to `compose()` and `link()`
- Fix: Optimize height of the legend
- Fix: Check if axes are enabled before updating them when the x or y scale changes
- Fix: Merge point selections on `SHIFT` instead of activating the lasso as `SHIFT` interferes with Jupyter Lab
- Fix: Allow to call `scatter.zoomTo()` with the same points multiple times
- Fix: Unfilter when calling `scatter.filter(None)`
- Fix: Properly listen to changes when setting custom `regl-scatterplot` options via `scatter.options()`

- Fix distributed build by ensuring that `jscatter/bundle.js` is included in the build
- Fix categorical encoding for partial data

> **Warning**: do not use this version! The distributed build is broken. Use `v0.12.6` instead. :pray:

- Ensure that the default point colors respect the background when setting both at the same time during initialization. I.e., in the following scenario, the point color will be set to _white_ by default as the background color was set to _black_:

py
jscatter.plot(data=df, background_color='black')


- Fix an issue when working with views of a pandas DataFrame where not all categorical data is present

- Loosen strictness of `rows` and `cols` of `compose()` to allow having empty cells in the grid