Replace dummy covariate and negative-control placeholders in `06_cm_spec.R`

Title
Replace dummy covariate and negative-control concept-set placeholders in cohort-method generation

Body

Problem

06_cm_spec.R still relies on placeholder handling for covariate and negative-control concept sets instead of real backed selections.

The generated script can therefore leave:

• includedCovariateConceptIds empty
• excludedCovariateConceptIds empty
• negative control inputs effectively unbound

Current behavior

The generator is compatible with HADES version skew, but the concept-set plumbing for covariates and negative controls is still placeholder-level rather than real workflow-backed selection.

Risk

• errorOnHighCorrelation = TRUE can fail without a meaningful exclusion path
• users may think concept-set-driven covariate selection is active when it is not
• generated cohort-method specs may be structurally valid but analytically incomplete

Requested change

Implement real handling for covariate and negative-control concept-set inputs in cohort-method generation.

Acceptance criteria

• covariate concept-set ids can be sourced from real workflow artifacts rather than dummy placeholders
• negative-control concept-set inputs are backed by actual selected/generated artifacts
• generated 06_cm_spec.R reflects real ids when present
• failure modes are explicit when required concept-set inputs are missing
• regression tests cover populated and unpopulated concept-set cases

Start in 06_cm_spec.R and trace how covariate and negative-control concept-set inputs are obtained from workflow artifacts. Verify the generated script with populated and unpopulated inputs, then add regression coverage showing real ids are reflected and missing required inputs fail explicitly.