Validate record fields of process record inputs at runtime

[bentsherman]

Close #7133

Type errors are only warnings in nextflow run (full type checking lives in nextflow lint, #7656), and isAssignableFrom treats any two record types as compatible. As a result, a process input declared with a named record type (sample: Sample) accepted any record at runtime, including ones missing required fields.

This PR checks each declared field of the record type when the input is bound, using the same rules destructured record inputs already follow:

• missing/null non-nullable field → task fails: input field `id` at index 0 cannot be null -- append `?` ...
• mismatched field type → warning naming the field: invalid argument type for input field `id` at index 0 -- expected a Boolean but got a String

Nested records are checked recursively. Type mismatches stay warnings because a hard error would give false positives at runtime (e.g. GString for String, Long for Integer).

Example (field world is missing):

record MyInput { greeting: String; world: String; language: String }

process HELLO {
    input:
    sample: MyInput
    ...
}

workflow {
    HELLO(channel.of(record(greeting: "world", language: "English")))
}

Before: task runs with world = null. After: the task fails with the error above.

Added a test to DataflowTypesTest. It fails without this change.

[netlify]

✅ Deploy Preview for nextflow-docs canceled.

Name	Link
🔨 Latest commit	074fd0b
🔍 Latest deploy log	https://app.netlify.com/projects/nextflow-docs/deploys/6abc09c6d869fa00081d2166

[adamrtalbot]

This is good for really early, defensive programming. Making sure no-one accidentally messes up the inputs. For a workflow:

nextflow.enable.types = true

record MyInput {
    greeting: String
    world: String
    language: String
}

process HELLO {
    input:
    sample: MyInput

    output:
    record(greeting: sample.greeting, language: sample.language, complete: file("output.txt"))

    script:
    """
    echo "${sample.greeting} ${sample.world} (${sample.language})" > output.txt
    """
}

workflow {
    sample = channel.of(
        record(greeting: "world", language: "English")
    )
    HELLO(sample)
}

N E X T F L O W  ~  version 26.09.1-edge
Launching `main.nf` [festering_varahamihira] - revision: b6e248a8c7
WARN: Type checking found 1 error(s) -- run `nextflow lint` to inspect them
ERROR ~ Error executing process > 'HELLO (1)'

Caused by:
  [HELLO (1)] input field `world` at index 0 cannot be null -- append `?` to the type annotation to mark it as nullable

Tip: you can replicate the issue by changing to the process work dir and entering the command `bash .command.run`

 -- Check '.nextflow.log' file for details

For AI, I expect this to be pretty powerful to allow it to iterate quickly.

The question is, is this too aggressive for some users? Forcing them to be very strict with records may slow them down.