feat: support every dialect sqlparser exposes

[funcpp]

Reconstructed from the dialect portion of #16, one of the contributions in #15.

What

Adds the 8 dialects sqlparser 0.62 exposes but this crate did not: duckdb, redshift, spark, clickhouse, sqlite, mssql, oracle, teradata. Dialect now covers all 16.

Name parsing moves into the core crate:

• impl FromStr for Dialect — canonical names plus the aliases postgres, sparksql, tsql, sqlserver; case-insensitive
• impl Display for Dialect, Dialect::name(), Dialect::ALL, Dialect::names()
• UnknownDialect error whose message lists every valid name
• Dialect becomes #[non_exhaustive] and gains PartialEq/Eq/Hash

The CLI and the Python binding now call str::parse instead of each keeping a copy of the name table.

Why

• The CLI, the Python binding, and the CLI's error string each held their own list, so adding a dialect meant editing four places. feat: expose lineage uncertainty and expand dialects #16 added its dialects to all of them again.
• #[non_exhaustive]: adding enum variants is already breaking, so this is the release to mark it — the next dialect addition then isn't.

Differences from #16

• Adds oracle and teradata, which feat: expose lineage uncertainty and expand dialects #16 missed. Both exist in sqlparser 0.62.
• Omits trino. feat: expose lineage uncertainty and expand dialects #16 mapped it to GenericDialect; sqlparser 0.62 has no Trino dialect, so that advertises support that isn't there and would change behavior silently if sqlparser later adds one. The README instead notes that generic parses Trino/Presto in practice.
• Name parsing is centralized rather than duplicated.
• Drops the aliases duck_db, click_house, ms_sql, t-sql, spark2 — they look derived from Rust variant names rather than real usage.
• Contains only the dialect change. feat: expose lineage uncertainty and expand dialects #16 also alters ColumnOrigin and ColumnLineage; those are separate decisions.

Tests

3 new, 91 total:

• every_dialect_parses_a_basic_query — every variant is wired to a working sqlparser dialect
• canonical_names_round_trip — name() and from_str are separate tables; this keeps them agreeing
• aliases_and_mixed_case_resolve

cargo fmt --all --check                                           exit 0
cargo clippy --workspace --all-targets --all-features -D warnings exit 0
cargo test --workspace --all-features                             exit 0

CLI checked by hand: -d duckdb, -d tsql, and -d presto (exits 1 with the full list).

Compatibility

Breaking for exhaustive downstream matches on Dialect — a minor bump (0.3.0). CLI behavior is unchanged: unknown names print to stderr and exit 1.

🤖 Generated with Claude Code