Benchmark submit advertises an unsupported preview flag

Status: open; help/parser consistency gap.

The operational CLI 1.2.1 advertises --dry-run for eval benchmark context submit, but the actual parser rejects it as an unrecognized argument before creating a job. Retained campaign evidence keeps that failed preview separate from the later confirmed submission.

Align command help with supported behavior. Either implement a side-effect-free preview that validates the exact job specification and execution owner or remove the unsupported flag from this command's help. Keep --confirm on submission.

The current campaign used an independently reviewed, explicit job JSON and a successful native preflight before its authorized --confirm submission. The first topology refusal was separately resolved by a private campaign topology declaring the local evaluation resource; deployed topology was unchanged.