enhancement: self-host compositional set helpers in stdlib

[vycdev2]

Summary

Move the compositional set helpers set.is_empty, set.union, set.intersection, and set.difference from hardcoded checker/interpreter cases into an exported compiler-shipped .jett set module while preserving their current generic signatures, membership results, and iteration-derived result order. Keep equality, storage, and cardinality operations as Rust-backed kernels.

Source documentation

• docs/progress.md — Phase K: Full Standard Library
• docs/progress.md — Phase D: Core stdlib
• docs/active/stdlib_json_extraction_plan.md — current compiler-aware builtin surfaces
• docs/architecture.md — Standard Library

Current state

The four helpers are implemented and behaviorally covered, but there is no stdlib/set.jett. Their public generic signatures remain hardcoded in jett_typecheck, and their bodies remain Rust match arms in jett_comptime.

The language can express this bounded slice compositionally:

• is_empty[T] can compare the retained set.length[T] kernel with zero.
• union[T] can start with the left set and add each right-side item through the retained set.add[T] kernel.
• intersection[T] and difference[T] can iterate the left set and select items through the retained set.contains[T] kernel.

The existing interpreter stores sets in insertion order. Source implementations should preserve the current observable order when results are later passed to set.to_list: union keeps left-side order before novel right-side items, while intersection and difference retain left-side order.

Scope

Includes:

• Add an exported namespace set stdlib source surface for is_empty, union, intersection, and difference.
• Remove only those four redundant checker signatures and interpreter bodies.
• Preserve generic inference, explicit type arguments, membership behavior, deduplication, and the current iteration-derived result order.
• Add regression coverage proving normal builds and symbol/signature queries resolve the migrated helpers from compiler-shipped set source.
• Keep standard-library progress, design, and architecture documentation aligned with the resulting boundary.

Explicitly excludes:

• Reimplementing the foundational set.new, set.add, set.remove, set.contains, or set.length equality/storage kernels.
• Migrating set.to_list or defining a new language-level set ordering guarantee.
• Changing set ownership, equality/hashability rules, generic inference, namespace/import rules, or the trusted stdlib-origin design tracked by design: define module imports and trusted stdlib origins #3.
• HIR, MIR, native runtime, or code-generation work.

Acceptance criteria

• Compiler-shipped stdlib source exports generic set.is_empty, set.union, set.intersection, and set.difference with their existing signatures.
• Empty, disjoint, overlapping, duplicate-add, and result-order cases pass for inferred and explicit generic calls.
• The four helpers no longer have hardcoded public signatures in jett_typecheck or Rust execution bodies in jett_comptime.
• set.new, set.add, set.remove, set.contains, set.length, and set.to_list remain explicit Rust-backed kernels.
• Project and dependency code still cannot reopen the builtin set namespace.
• A focused query or loader regression identifies the compiler-shipped set source as the definition location for at least one migrated helper.
• docs/design.md, docs/architecture.md, and docs/progress.md distinguish source-hosted set helpers from retained kernels.
• cargo fmt --check, cargo build, and cargo test -q pass.

Dependencies / open questions

Coordinate builtin-type namespace handling with PR #56 and issue #57 so string, list, and set use one narrow compiler-shipped rule rather than parallel special cases. This work must not broaden or resolve the module/import/trusted-origin policy in #3.

This was generated by an AI agent (vycdev2). Please verify any changes before merging or applying.