Folio rename decl/rename tests never exercise a non-contact-length leaf

[mikebronner]

Summary

Implements #191 — closes the span-length blind spot in the Folio rename test suite.

Every test that drives the production decl/rename logic seeded a contact leaf (7 chars), so the re-read name('...') span always coincided with the NAME_START/NAME_END constants. A span-computation regression for a different name length would have passed green. This adds two tests that drive the same production functions against a 9-char dashboard leaf and assert the literal columns the page body dictates.

Changes

• Add decl_range_at_returns_the_page_name_span_for_a_non_contact_length_leaf — drives decl_range_at with a dashboard leaf, asserting the prepare_rename Range covers columns 12..21 (quote-excluded), not the constants.
• Add collect_targets_rewrites_a_non_contact_length_leaf_declaration — drives collect_route_declaration_targets with the same leaf, asserting start_column == 12 and end_column == 21 on the returned EditTarget.
• No change to seed_page: it already derives the body's name('<leaf>'), filename, and uri from the leaf (merged PR seed_page still hardcodes contact-specific page content and uri that don't track route_name #190 / issue seed_page still hardcodes contact-specific page content and uri that don't track route_name #188), so it produces a consistent dashboard fixture for any length and all existing contact-seeding tests pass unchanged.

Acceptance Criteria

• seed_page writes <?php name('<leaf>'); ?> derived from the actual leaf segment, so body and route-index key stay consistent for any leaf length; existing contact/admin.contact tests pass unchanged. (Already satisfied by PR seed_page still hardcodes contact-specific page content and uri that don't track route_name #190 — verified, no change needed.)
• A new #[tokio::test] drives decl_range_at with a non-contact-length leaf (dashboard, 9 chars), asserting the returned Range covers columns 12..21, NOT the NAME_START/NAME_END constants.
• A new #[tokio::test] drives collect_route_declaration_targets with the same leaf, asserting start_column == 12 and end_column == 21 on the returned EditTarget.
• cargo test -p laravel-lsp folio_rename passes with all tests green (14 tests: 2 new + 12 pre-existing).

Test Plan

• cargo test -p laravel-lsp folio_rename — 14 passed, 0 failed.
• cargo fmt --check — clean.
• cargo clippy --tests -p laravel-lsp — no new lints.

Fixes #191

[mr-sherlock-holmes]

✅ Approved

Review Summary

• Test-only PR (+86 −0): two new #[tokio::test]s in laravel-lsp/src/tests/folio_rename.rs that drive the Folio decl/rename logic against a 9-char dashboard leaf instead of the 7-char contact leaf every other test uses — closing the span-length blind spot #191 flagged.
• AC #1 — met (by existing state, noted divergence). The AC anticipated seed_page being updated to derive the body from the leaf; that update already landed in PR #190 (closing #188). seed_page writes format!("<?php name('{leaf}'); ?>") (folio_rename.rs:91) from the leaf at :88, so no change was needed here — the new tests confirm it produces name('dashboard') for a dashboard leaf, and the pre-existing contact/admin.contact tests stay green. Intent satisfied; flagging the divergence for the record.
• AC #2 — met. decl_range_at_returns_the_page_name_span_for_a_non_contact_length_leaf asserts the returned Range is start character: 12, end character: 21 as hardcoded literals, not NAME_START/NAME_END. Column math checks out: in <?php name('dashboard'); ?> the 9-char content sits at cols 12..=20, quote-excluded span 12..21.
• AC #3 — met. collect_targets_rewrites_a_non_contact_length_leaf_declaration asserts t.start_column == 12 and t.end_column == 21 (literals), plus targets.len() == 1, file_path, line == 0, and new_text == "home". Call convention matches the sibling collect_targets_rewrites_the_page_name_declaration.
• AC #4 — met. CI green (LSP — test, fmt, clippy passed), and the review lenses independently ran cargo test -p laravel-lsp folio_rename — all 14 tests pass, including the two new ones.

The tests are honest regression guards: because they assert the literal 12/21 columns rather than the contact-sized constants, a span-computation bug for any other name length would fail red — exactly the gap #191 set out to close.

📋 Non-blocking follow-ups

• None.

Ready for @mikebronner to merge.