Closes part of #7202.

The task cache hash keys on environment module directives (TaskHasher adds task.config.getModule() to the hash), but the lineage TaskRun record had no field for them. As a result, comparing two task runs with nextflow lineage diff could not surface a module change that invalidated the cache, even though the record already captures the other environment inputs (conda, spack, container, architecture).

This adds a module field to the TaskRun v1beta1 model and populates it in LinObserver, alongside the existing environment fields.

Example

process GREET {
    module 'samtools/1.9'
    input:  val name
    output: path 'out.txt'
    script: """ echo "Hello ${name}" > out.txt """
}
workflow { GREET(Channel.of('world')) }

Changing samtools/1.9 to samtools/1.17 and re-running with -resume correctly re-runs the task. Before this change, lineage diff of the two task runs showed only the parent run id changing; now it shows the module field changing.

Scope

This covers the module directive only. The eval output commands and the stub-run flag (the other cache-hash inputs missing from the record, per #7202) are left out: representing eval commands needs a design decision, and the stub flag is partially covered already via codeChecksum switching to the stub source. Those remain tracked in #7202.

This makes the record consistent with that cache-hash input; it does not make lineage diff a full substitute for -dump-hashes, since the record stores resolved values rather than the ordered hash-key list and DataPath checksums always use the default hash mode.

Tests

• LinEncoderTest now round-trips a populated module field.
• LinObserverTest, LinCommandImplTest updated for the new field.
• LinTypeAdapterFactoryTest confirms records without a module field still decode (backward compatible).