feat: map lambda.z.corrxy to the CDISC CORRXY term

[billdenney]

Part of #1102; the CORRXY item on the 0.12.2 checklist in #481.

PKNCA 0.12.1.9000 reports lambda.z.corrxy, the correlation between time and log-concentration for the terminal slope. aNCA had no CDISC entry, so it came through unlabelled and the expected-label test in test-pivot_wider_pknca_results.R failed.

Adds the row to data-raw/metadata_nca_parameters.csv and regenerates data/metadata_nca_parameters.rda. Classified with the other half-life diagnostics (unitless, Standard, Half-life, depends on half.life), matching how r.squared and adj.r.squared are already handled.

[Gero1999]

Thanks for this — the CDISC mapping itself is correct and I verified the rebuilt .rda matches the CSV (all 141 params present, lambda.z.corrxy -> CORRXY with the right PPTEST). Classifying it with the other half-life diagnostics (depends="half.life", unitless, Half-life) is consistent with how r.squared/adj.r.squared are handled.

One concern that needs aNCA-specific context before merge — the test change in test-pivot_wider_pknca_results.R:

aNCA's CI installs PKNCA from CRAN (0.12.1) — DESCRIPTION pins PKNCA (>= 0.12.1) and there is no Remotes: entry, so setup-r-dependencies pulls 0.12.1, not the dev build. In 0.12.1, lambda.z.corrxy does not exist (git show v0.12.1:R/half.life.R has zero matches), so pk.nca() never emits a CORRXY column. The expected_labels assertion is an exact ordered expect_equal, so adding CORRXY makes it fail on CI until 0.12.2 is on CRAN:

• dev PKNCA: CORRXY present -> label needed -> passes ✅
• CRAN 0.12.1 (what CI runs): CORRXY absent -> extra label -> expect_equal fails ❌

This is the mirror image of the R2/R2ADJ labels already in that vector: those exist in 0.12.1 too, so they're safe today. CORRXY only exists in dev.

The CDISC data change (CSV + .rda) is safe to merge as-is — an unused extra metadata row is harmless on 0.12.1. It's specifically the test assertion that's coupled to the dev PKNCA output.

Options (any is fine, your call):

1. Make the CORRXY label conditional on it being present in the output (e.g. only assert it when "CORRXY" %in% names(labels)), so the test passes on both PKNCA versions.
2. Split: land the CSV/.rda now, hold the test assertion until aNCA moves its minimum to the PKNCA release that ships lambda.z.corrxy (add Remotes: humanpred/pknca or bump the floor).
3. If the intent is that this whole branch targets 0.12.2 and CI will be pointed at dev PKNCA before merge, then it's fine as-is — just flagging so it's a deliberate decision rather than a surprise red check.

Everything else here looks good. Thanks again!