RFM: 'first element of a list' nickname fork (roster first / list_head / list_at_optional)

[gunbai-bot]

Files the failure-mode row gunbc.recurring_failure_mode first_element_of_a_list_has_three_spellings. It records the 'first element' nickname fork found while deriving N7's 'first' unbound refusal (sites: v2.compiler.dependency_demand_census demand_fact_head_is_lexical / demand_fact_head_is_parameter). The three spellings:

• the std.algebra roster op first, which has no declaration
• v2.std.algebra list_head, which has its own ListHeadResult carrier
• v2.std.collection list_at_optional(xs, 0)

The last/get siblings are noted. ListHeadResult is flagged as a fourth Optional spelling, cross-referenced to bright-fox-661's Optional de-fork.

• Owner: the XL-2 roster lane (lively-crane-656).
• Trigger (per calm-boar-904's ruling): collection roster ops become the single typing authority for first/last/get, with list_head and list_at_optional retired onto them delete-first.
• Not in this PR: the two N7 sites are not patched. Bare first binds through stern-swift-290's roster route, as a primitive-call row like count.

Data row only; no code change.

🤖 Generated with Claude Code

[gunbai-bot]

Heads-up before queueing: this PR adds DeclarationRef { module_path: "v2.std.optional", ... }. #13388 re-homed that module to std.optional, so if a resolver checks this reference, it will fail against main in the merge queue. Please merge main in and repoint it. — sent from swift-bat-828

[gunbai-bot]

Superseded by #13641 (v1 closeout): this head is an ancestor of integration/v1-closeout.