Backport #309 to maint/v3 (3.5.4)

[ehrlinger]

Backports #309 to maint/v3, ahead of the 3.x release. 3.5.4 is unreleased (CRAN has 3.5.3), so the NEWS bullets go under 3.5.4 and the version stays put.

What it fixes in gg_partial_rfsrc() / plot.gg_partial_rfsrc():

• partial.type = "mort" stopped with "replacement has 3 rows, data has 23" under the default partial.time. Mortality is summed over every event time, so its data now carry no time column, and both panels are labelled "Predicted Mortality".

• On a survival forest with xvar2.name, the continuous panel now gives each level its own panel, one line per time horizon, instead of one zigzag per time.

• A factor xvar2.name is passed to partial.rfsrc() as integer codes, and grp comes back as a factor holding the labels.

• Factor levels are now coded by the fitted forest, not by newx (2fcf7d0a, raised by Copilot on fix: mortality, grouped survival lines and factor xvar2 in gg_partial_rfsrc #309). A factor re-levelled in newx had passed one level's code under the other's label, for xvar.names and xvar2.name alike. A level the forest was not trained on is now an error.

Adapted for maint/v3

This is a cherry-pick of 9ca657b2 and 2fcf7d0a (-x). The second applied clean; the first needed these conflicts resolved:

• v3's plot.gg_partial_rfsrc() has no labels = argument, so the shared partial_rfsrc_facet() helper takes no strip labeller here, and the facets match v3's existing calls.
• The xvar2.name roxygen keeps v3's wording and adds only the factor sentence. Main's longer newx text came from a separate change that was never backported.
• The plot.gg_partial_rfsrc: categorical partial dependence stacks per-observation predictions instead of averaging #299 NEWS bullet keeps v3's wording; the three new bullets sit under it in 3.5.4.

Verification (on this branch)

• devtools::document(), then lintr::lint_package(): 0 lints
• NOT_CRAN=true VDIFFR_RUN_TESTS=true devtools::test(): FAIL 0, SKIP 5, PASS 1615. The new gg-partial-rfsrc-continuous-survival-by-group baseline from main matches v3's rendering unchanged; 52 baselines before and after.
• R CMD check --as-cran with the manual, from a git archive export: 1 NOTE (CRAN incoming feasibility). Timed steps total 213 s locally. Tarball 2.4 MB, with no hidden files other than .Rinstignore.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 97.56098% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 89.40%. Comparing base (0cbe923) to head (1a4440d).
⚠️ Report is 1 commits behind head on maint/v3.

Files with missing lines	Patch %	Lines
R/gg_partial_rfsrc.R	95.65%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           maint/v3     #312      +/-   ##
============================================
+ Coverage     89.12%   89.40%   +0.27%     
============================================
  Files            50       50              
  Lines          4682     4699      +17     
============================================
+ Hits           4173     4201      +28     
+ Misses          509      498      -11     

Files with missing lines	Coverage Δ
R/plot.gg_partial.R	95.23% <100.00%> (+10.43%)	⬆️
R/gg_partial_rfsrc.R	90.19% <95.65%> (-0.32%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ehrlinger]

Added 2fcf7d0a from #309, the fix Copilot's review there prompted: factor levels are now coded by the fitted forest, not by newx. Without it this backport would have shipped the xvar2.name factor path with the same level-swap bug. It applied clean, and its NEWS bullet is copied from main into 3.5.4.

Re-run on 1a4440d3: 0 lints; NOT_CRAN=true VDIFFR_RUN_TESTS=true devtools::test() FAIL 0, SKIP 5, PASS 1615; R CMD check --as-cran with the manual, from a git archive export: 1 NOTE (incoming feasibility). 52 baselines before and after.