Fix Inf coordinate values causing shapes not to render

[Chr96er]

Summary

Fixes #2364 - JSON doesn't support Inf values, causing geoms to not render when users specify Inf/-Inf coordinates (a common pattern for "extend to edge" annotations).

Changes

• Added centralized replace_inf_in_data() function in R/utils.R that replaces Inf values with panel limits
• Applied after to_basic() returns in R/layers2traces.R, handling all geoms generically (geom_line, geom_rect, geom_polygon, geom_ribbon, geom_segment, etc.)

Test plan

• Added test for geom_line with Inf y values (original issue)
• Added test for geom_rect with Inf x/y values
• Added test for geom_polygon with Inf x/y values
• Added test for faceted plots with Inf values
• Tests verify replaced values match expected panel limits
• All 211 relevant tests pass

[cpsievert]

Thanks for this contribution, @Chr96er! This is a really nice approach — handling Inf replacement centrally after to_basic() returns is clean and covers all the geoms generically. This is very close to being ready to merge. A few things I'd like to see addressed:

NEWS.md entry

Please add a bullet under the # plotly (development version) section in NEWS.md, e.g.:

* Fixed Inf/-Inf coordinate values (e.g., in `geom_rect()`, `geom_ribbon()`, `annotate()`) causing shapes not to render (#2364).

Comment accuracy in replace_inf_in_data()

The comment says it "handles x, y, xmin, xmax, xend, ymin, ymax, yend columns generically", but after to_basic() runs, nearly all geoms have already converted those columns to just x/y (e.g., to_basic.GeomRect converts xmin/xmax/ymin/ymax → x/y, to_basic.GeomSegment converts xend/yend → x/y, ribbon_dat() converts ymin/ymax → y). The extra column checks are still useful for geoms that go through to_basic.default (which returns data as-is), like GeomTile with continuous fill. Could you update the comment to reflect this more precisely?

Test coverage for geom_ribbon and geom_segment

The PR body mentions these geoms as fixed, and I've verified they work, but there aren't tests covering them directly. Adding a test or two for geom_ribbon(aes(ymin = -Inf, ymax = y)) and/or geom_segment with Inf endpoints would help prevent regressions.

Minor formatting nit

In test-ggplot-rect.R around line 218, the for loop has inconsistent indentation:

 for (tr in rect_traces) {
    expect_false(any(is.infinite(tr$y), na.rm = TRUE))

The for line has one space of indent instead of two.