fix(brain): make automatic feedback explicit and target-safe

[ohdearquant]

Summary

• stamp every new explicit-feedback event with an internally selected originating_verb, while keeping the canonical stored event verb stable for replay and historical queries
• expose per-origin feedback counts with a legacy fallback so brain.feedback, brain.auto_feedback, and the deprecated brain.emit path can be measured independently
• make omitted brain.auto_feedback.signal an explicit abstention that writes no feedback event and performs no posterior update
• require signaled automatic feedback to name one unique results[].id, use that selected result for target and serve attribution, and retain the full result set only as candidate context
• align the public help, operator guidance, skills, and ADRs with the non-forgeable provenance and abstention contracts

Why these issues belong together

#1587 is the observability prerequisite for #1588: automatic-feedback behavior cannot be evaluated if its emitted events remain indistinguishable from direct feedback. This PR first preserves the originating path on the event, then removes the rank-first positive-feedback loop behind a target-safe contract that can be measured by that provenance.

Verification

All checks passed at exact commit 3f3687b01bbf8b0740d71694f0ff2bbd27cd92b6:

• cargo test --manifest-path crates/Cargo.toml --workspace
• cargo check --manifest-path crates/Cargo.toml --workspace
• cargo clippy --manifest-path crates/Cargo.toml --workspace --all-targets -- -D warnings
• cargo fmt --manifest-path crates/Cargo.toml --all -- --check
• RUSTDOCFLAGS=-Dwarnings cargo doc --manifest-path crates/Cargo.toml --workspace --no-deps

An independent review found no actionable issues.

Closes #1587
Closes #1588

AI-assisted contribution: Codex prepared this change and PR description.