Add testthat coverage for trait meta-analysis pipeline functions

[omkarrr2533]

What this does

Adds test coverage for the core functions in the meta-analysis pipeline — check_consistent(), run.meta.analysis.pft(), meta_analysis_standalone(), p.point.in.prior(), and documents a structural issue in get.parameter.samples().

These tests lock in existing behavior so that future refactoring (particularly the GSoC modularity work) has a safety net against regressions.

Why

The meta-analysis pipeline currently has minimal test coverage. Before changing anything in these functions, we need tests that confirm what they actually do today. A few things I found worth noting while writing these:

• get.parameter.samples() ends with save(...) and no return() statement, so it returns invisible(NULL). Callers can't access results without loading the saved file. The test file documents this with a skip explaining the problem — this is the core architectural issue the GSoC project aims to fix.
• run.meta.analysis.pft() checks for input files before touching the database, which means we can test all its precondition logic without needing a DB connection.
• check_consistent() is a pure function with no side effects, so it's straightforward to test thoroughly.

What's covered

Function	Tests	What they check
check_consistent()	7	Normal and gamma priors, both tails, custom thresholds, input validation
run.meta.analysis.pft()	9	Missing file errors, skip-when-cached logic, empty trait data, string coercion, posteriorid checks
meta_analysis_standalone()	2 (skip if no JAGS)	Greenhouse data filtering, full return structure verification
p.point.in.prior()	4	Normal/gamma distributions, extreme values, return type
get.parameter.samples()	0 (1 documented skip)	Documents the invisible(NULL) problem structurally

22 tests run, 1 strategic skip.

Design decisions

• No database needed. Every test either targets precondition checks that fire before any DB call, or uses the shared fixtures from helper-test-data.R. The dbcon = NULL calls are safe because run.meta.analysis.pft() validates input files before attempting any database operations.
• JAGS tests skip gracefully. The integration test in test-meta_analysis_standalone.R checks for rjags installation and actual JAGS availability before running.
• Shared fixtures in helper-test-data.R. create_test_trait_data(), create_test_priors(), create_test_pft(), and setup_trait_files() build minimal but realistic data structures that match what the pipeline actually produces. These are reusable for future tests.
• Uses ::: for internal functions. check_consistent, p.point.in.prior, and run.meta.analysis.pft aren't exported, so tests access them via PEcAn.MA:::.

Checklist

• All tests pass locally with devtools::test()
• No changes to existing source code
• Uses skip_on_cran() and skip_if_not_installed() where appropriate
• R CMD check passes without new warnings