fix(ogar-adapter-clickhouse-ddl): preserve backtick-quoted dotted table names on parse

[AdaWorldAPI]

Summary

Closes Codex P2 on PR #38 (now merged) — last_ident split on . BEFORE stripping backticks, silently truncating `sale.order` to order and breaking the round-trip the PR claimed to support for Odoo dotted names. Per your instruction, fresh PR for visibility since #38 is on main.

PR #39 had no review comments — only #38 needed a fix.

The bug

ct.name (ObjectName) ──to_string──▶  "`sale.order`"
                     ──last_ident──▶  rsplit('.').next() = "order`"
                                      ──trim_matches('`')  = "order"
                                                            ^^^^^^^ WRONG

The . was treated as a database/table separator even though it was inside the quoted identifier. Result: emit + parse round-trip turned sale.order into order, breaking the round-trip the PR claimed to support for Odoo dotted names.

The fix

Bypass the string round-trip entirely. sqlparser already resolves the backtick/dot ambiguity structurally:

Input	ObjectName shape
`sale.order`	ONE part: Ident { value: "sale.order", quote_style: Some('') }` — backticks group it
db.table	TWO parts: [Ident{value:"db"}, Ident{value:"table"}] — . is the separator

New helper accesses the last part's Ident.value directly (the unquoted text):

fn last_ident_from_object_name(name: &ObjectName) -> String {
    name.0.last()
        .and_then(|part| part.as_ident())
        .map(|ident| ident.value.clone())
        .unwrap_or_default()
}

Correct for both cases without any string splitting or backtick-trimming heuristics.

Verifying tests (4 new; 12 prior → 16 total)

Test	Asserts
parse_backtick_quoted_dotted_table_keeps_full_name	`sale.order` → name == "sale.order" (not "order" — the Codex P2 motivating case)
parse_qualified_db_table_takes_last_segment	my_db.users → name == "users" (qualified-db path still works)
round_trip_odoo_dotted_table_name	Class::new("sale.order") + attribute → emit → parse → name + attribute byte-identical
round_trip_three_segment_odoo_name	account.move.line three-segment name round-trips

Verification

$ cargo test -p ogar-adapter-clickhouse-ddl                            -> clean
$ cargo test -p ogar-adapter-clickhouse-ddl --features clickhouse-parser
                                                                       -> 16/16
$ cargo check --workspace --all-targets                                -> clean

PII abort-guard (word-boundary): CLEAN.

Symmetry note

This is the same class of bug ogar-adapter-surrealql shipped a fix for in PR #36 (backtick-quoting on emit + symmetric parser handling). The two adapters now follow the same discipline:

• surrealql adapter: emit backtick-quotes non-bare idents; walker accesses Path.start.text directly (unquoted text) — no string round-trip on the parse side.
• clickhouse adapter: emit backtick-quotes non-bare idents; walker accesses ObjectNamePart::Identifier.value directly (unquoted text) — no string round-trip on the parse side.

Same lesson: don't string-round-trip when the AST already encodes the structural distinction. Worth pinning in a future ADR if a third adapter trips the same wire (the ogar-adapter-ttl adapter doesn't have this risk because Turtle has different identifier semantics).

Out of scope for this PR

The DOMAIN-INSTANCES.md §2.6 + RDF-OWL-ALIGNMENT.md §10 Phase 2c updates I was queuing for after lance-graph PR #473 lands — those are now also unblocked and will be a separate small docs PR. Per the user's instruction to ship the fix PR first for visibility, the docs PR follows.

https://claude.ai/code/session_01PBTGaPCSnnt6u3pjXpbLwY