Fix summary-exclusion flags filtering individual plots (#1438)

[Gero1999]

Issue

Closes #1438

Description

Summary-exclusion flags (PKSUMXF for ADNCA, PPSUMXF for ADPP) were applied to every TLG graph, so individual concentration plots (pkcg01) and combined plots (pkcg02) hid summary-excluded records. The documented intent (#1018, #1436) is that these flags remove records from summary tables and mean plots only, while keeping them everywhere records are inspected individually.

The old routing (tlg_data_key()) was a binary listing-vs-everything switch: listings got unfiltered data, everything else (all tables and all graphs) got filtered data. That scope quietly widened from "mean plots" to "graphs" in #1356.

This inverts the default: only summary/mean functions self-filter, and the filtered/_all boundary split is retired so there is a single source of truth per dataset.

Changes:

• Add filter_summary_excluded() in R/utils-tlg.R. It auto-detects the relevant flag (PKSUMXF/PPSUMXF), is a no-op when absent, scopes each dataset to its own flag (no cross-filtering), and uses dplyr::filter() so column label attributes survive the row subset.
• Call it at the top of the summary/mean functions only: t_pkct01, t_pkpt03_col, t_pkpt07_norm, t_pkpt08_uri, t_pkpt11_gmr, pkcg03 (mean conc), and p_pkpg01_cum/p_pkpg02_doseprop/p_pkpg03_boxp. Their delegators (_dose, _tad, _MP_col, p_pkpg01_per, p_pkpg04/06, g_pkcg03_*, p_pkcg03_*_dose) inherit automatically.
• Individual (pkcg01) and combined (pkcg02) plots and all listings now receive raw data.
• Retire tlg_data_key() and filter_tlg_excluded(); tab_tlg.R exposes one unfiltered source per dataset.

Because filtering now lives in the TLG functions, the exported R script path gets the correct behaviour too.

Combined plots (pkcg02) decision: kept unfiltered. pkcg02 overlays individual subject traces, so records are still being inspected individually.

Unrelated naming bug found while testing (#1448): adding name-based assertions surfaced a pre-existing bug on main — pkcg01/pkcg02 named their output list from adnca[["id_plot"]], but id_plot only exists on the intermediate adnca_grouped, so the list came back unnamed. No prior test checked plot names, so it went unnoticed. Fixed here by naming from adnca_grouped (tracked separately in #1448).

Definition of Done

• Individual plots show summary-excluded records
• Decision made for combined plots (pkcg02: unfiltered)
• Summary tables and mean plots still filter
• Listings unchanged (unfiltered)
• Per-dataset scoping preserved (no cross-filter)
• Works in the exported R script path (functions self-filter)
• Misleading "mean plots" routing comments corrected
• Tests assert individual-retains vs summary-drops

How to test

Flag a record with PKSUMXF == "Y" (ADNCA) or PPSUMXF == "Y" (ADPP) and confirm in the TLG tab that:

• it still appears in individual (pkcg01) and combined (pkcg02) plots and in listings, and
• it is dropped from summary tables (t_pkct01, t_pkpt*) and mean plots (pkcg03, p_pkpg*).

Contributor checklist

• Code passes lintr checks
• Code passes all unit tests
• New logic covered by unit tests
• New logic is documented
• App or package changes are reflected in NEWS
• Package version is incremented
• R script works with the new implementation (if applicable)
• Settings upload works with the new implementation (if applicable)
• If any .scss change was done, run data-raw/compile_css.R
• If a package dependency was added/changed, run data-raw/test_suggests_hidden.R

Notes to reviewer

R is not available in this environment, so tests/lintr were verified statically (line-length and brace/paren balance) rather than executed; please run devtools::test() and lintr::lint_package() locally.

main has been merged into this branch and conflicts in inst/shiny/modules/tab_tlg.R and tests/testthat/test-tab_tlg.R resolved (kept the single unfiltered ADPP source from this branch and folded in the ADPP "run NCA first" notification from #1335).

Two test failures from the first CI run are addressed: the g_pkcg01 name assertion failed due to the #1448 naming bug (fixed here), and the t_pkct01 exclusion test expected an n == 0 row for a fully-excluded cell — the table drops empty groups, so the test now asserts the row is absent.