Validate version argument (fix NA crash)

[karstenm1987]

Tracked upstream at Global-Macro-Database-Internal#413. Same filter applied to the recent Stata PR: fix actual bugs, skip nice-to-haves.

Bug

gmd(version = NA) crashed with the cryptic R-level error "missing value where TRUE/FALSE needed". Root cause at R/gmd.R:263:

if (!is.null(version) && tolower(version) == "list") { ... }

With version = NA, tolower(NA) returns NA, NA == "list" returns NA, and if (NA) is an error. Same failure mode for version = 2025 (numeric) or a multi-element character vector.

Fix

Reject those inputs at the top of the version-check block with a clear message that names the argument.

Verified

> gmd(version = NA,               country = "USA", variables = "rGDP")
Error: `version` must be a single non-NA string (e.g. "2025_09"), "current", or "list".
> gmd(version = NA_character_,    country = "USA", variables = "rGDP")
Error: `version` must be a single non-NA string (e.g. "2025_09"), "current", or "list".
> gmd(version = c("2025_09","2026_03"), country = "USA", variables = "rGDP")
Error: `version` must be a single non-NA string (...).
> gmd(version = 2025,             country = "USA", variables = "rGDP")
Error: `version` must be a single non-NA string (...).

# Regressions — all still work:
> gmd(version = "2025_09", country = "USA", variables = "rGDP")   # 242 rows
> gmd(version = "current", country = "USA", variables = "rGDP")   # 241 rows
> gmd(version = "list")                                            # prints list
> gmd(country = "USA", variables = "rGDP")                         # 241 rows (NULL default)

Out of scope

Other issues catalogued in #413 (double "Error:" prefix, case-sensitivity inconsistency across country/variables/sources, iso=TRUE warning vs error, empty-vector handling, dependency cleanup, start_year/end_year not implemented) are UX/consistency issues rather than functional bugs, kept for separate PRs.

Also note: the listed-but-404 versions on S3 (2025_01/03/05/06/08) are a release-pipeline issue and not fixable here.