Pointer events in the pixel-centre convention; bold and outlined text markers

[CSSFrancis]

Closes #72. Closes #66.

#72 — pointer coordinates are half a pixel off

The same physical click (cyan cross) recorded from the main renderer and from this branch. The red square is the pixel round(event.img_x), round(event.img_y) picks.

_imgToCanvas2d puts image coordinate i at the centre of pixel i, the same convention marker offsets, widget positions and Python's display_to_data use. _canvasToImg2d skipped the matching -0.5, so the two weren't inverses. Every 2-D pointer event read half a pixel right and down of anything drawn at the same spot.

• _canvasToImg2d now subtracts the 0.5. The round trip is exact at every zoom and pan (it was off by 0.5 before).
• Two small helpers keep the convention in one place: _imgPix2d (floor(v + 0.5)) gives the pixel index and _inImgAxis2d checks bounds. The readout, the value probe, the brush bounds and _inImage2d all use them. _pixelValue2d adds the 0.5 back itself, because it addresses texels and the detail tile by pixel edge.
• xdata/ydata for pointer_down, double_click, pointer_settled, key_down and the readout now come from _imgToAxisVal2d:
  • imshow (one axis value per pixel centre): pixel i reads exactly x_axis[i]. Past the first or last pixel centre the value extends linearly instead of clamping.
  • pcolormesh (cell edges): uses (i + 0.5) / n, which gives exactly the values it reported before, so mesh xdata doesn't move.
• The cursor-anchored wheel zoom adds the 0.5 back, so zooming still keeps the image point under the cursor fixed.
• Side effect: brush strokes now land under the cursor instead of half a pixel up-left of it.

This is a behaviour change, so it has an api_change fragment. Handlers should use round(event.img_x) to get a pixel index. int() no longer does that, and docs/events.rst used it, so the docs are updated. The bundled examples already used round().

Not changed: the imshow tick gutters still stretch x_axis[0]…x_axis[-1] from the image's left edge to its right edge, not from the first pixel centre to the last. A tick label can therefore sit up to half a pixel from the xdata of a click at the same spot. That mapping was already inconsistent with marker placement. Fixing it moves tick positions, so it is a separate stacked PR: #74. That PR also turned up a letterbox bug in the gutters and the missing origin='lower' y ticks.

#66 — bold and outlined text markers

add_texts (Plot2D and Plot1D) and add_text gain three options:

• fontweight: "normal" (default), "bold", or a CSS numeric weight in [1, 1000]. The name matches fontsize, following matplotlib.
• outline_color: a halo colour, None by default. The halo is drawn with strokeText under fillText, with round joins.
• outline_width: default 3.0. Roughly half of it shows outside the glyphs.

The options are validated on creation and on MarkerGroup.set, so a bad value is refused before it reaches the group. The 1-D and 2-D renderers share _markerTextStyle / _drawMarkerText. Plot2D's add_texts also gets the docstring it was missing.

plot.add_texts([[40, 12]], ["2 Å"], fontsize=16, fontweight="bold",
               color="#ffffff", outline_color="#000000")

Tests

• test_plot2d/test_pointer_pixel_centre.py: round trip of the extracted JS helpers across zoom and pan, the imshow and mesh axis mapping, and real Playwright clicks placed via data_to_display. The three click tests fail on main and pass here.
• test_markers/test_text_style.py: wire format, validation, live set, and screenshot checks on 1-D and 2-D panels. Bold puts down more ink; the outline paints its own colour while the fill stays on top.
• FIGURE_ESM.md anchors are updated to the current line numbers. They were uniformly one line off on main.
• Full suite run locally (Windows, Chromium): 2438 passed, 58 skipped.

https://claude.ai/code/session_01EUrvzeXdzNjPKBTk5jCtp1

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.16%. Comparing base (ce451b6) to head (6c93609).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #73      +/-   ##
==========================================
+ Coverage   91.12%   91.16%   +0.03%     
==========================================
  Files          41       41              
  Lines        4847     4865      +18     
==========================================
+ Hits         4417     4435      +18     
  Misses        430      430              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.