fix(utils): accept NULL as synonym for NA in read.output start/end year

[anshul23102]

For non-difftime inputs, start.year and end.year used NA as the sentinel for "read all years", while variables, ncfiles, and pft.name all used NULL. This inconsistency was confusing and undocumented.

NULL is now silently converted to NA at the top of read.output() so both sentinels work identically. Default values and all existing behavior are unchanged.

Motivation and Context

Fixes #3864, flagged by @infotroph. The inconsistency made it easy to accidentally pass NULL expecting "read all years" and get unexpected behavior instead. Aligning the sentinel values makes the API consistent with every other parameter in the same function.

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• All new and existing tests passed
• I agree that PEcAn Project may distribute my contribution under BSD 3-clause license
• I have read the CONTRIBUTING document

[anshul23102]

Summary of changes

This PR fixes issue #3864 by accepting NULL as a synonym for NA in the start.year and end.year parameters of read.output().

What was changed

base/utils/R/read.output.R

• Added two lines at the top of the function body that convert NULL to NA for start.year and end.year. This is consistent with how other parameters in the same function (variables, ncfiles, pft.name) use NULL to signal "use all".
• Updated the @param documentation to mention both NA and NULL as accepted sentinel values.

base/utils/man/read.output.Rd

• Synced the generated man page with the updated roxygen documentation (added "or NULL" to the start.year/end.year parameter description).

base/utils/tests/testthat/test-read.output.R

• Added a new test_that block that verifies start.year = NULL, end.year = NULL gives the same result as start.year = NA, end.year = NA.
• The test uses its own isolated nulldir <- withr::local_tempdir() rather than the shared testdir defined at the top of the file. The shared testdir accumulates .nc files across test blocks, so using it would give 3 years of data instead of the 2 that the test creates, causing a spurious failure.

CI notes

The check (check_base, ...), check (check_models, ...), and extras failures are pre-existing infrastructure issues unrelated to this PR:

• The extras Docker build failures have been present across many PRs (e.g., input.ens.gen: Parent handling improvements #3983, fix: replace 1:nrow anti-pattern with seq_len(nrow()) in benchmark mo… #3984, Make trait_names optional in get_trait_data_pft() #3985).
• The check_base failures are caused by libsodium-dev not being installed on the CI runner when it tries to build modules/data.atmosphere as a transitive dependency. The same failure appears on other open PRs (e.g., add CalAdapt WRF download function for CMIP6 hourly met data #3967) that touch unrelated code.

All base/utils tests pass: [ FAIL 0 | WARN 0 | SKIP 1 | PASS 232 ] on both R 4.2 and R 4.4.

[infotroph]

Thanks! Two quick tweaks in the inline comments, and can you please also add a bullet in NEWS.md?

[anshul23102]

Done - applied all three suggestions:

• Default changed from NA to NULL for both start.year and end.year
• Doc simplified to just say If \NULL`(dropping theNA` mention)
• Added a ## Changed bullet to NEWS.md

Both NULL and NA still work identically internally; NULL is now the advertised default.