feat: add oracle_sql dialect support

[julien-pinchelimouroux]

Add basic support for Oracle SQL dialect to force quotes on identifiers.

[prql-bot]

The cli::test::get_targets snapshot will fail in CI. #[strum(serialize_all = "lowercase")] on Dialect lowercases variants without adding separators — OracleSql serializes as oraclesql, not oracle_sql. Running prqlc list-targets against this branch locally outputs sql.oraclesql, so the expected line added at cli/test.rs:89 won't match.

That also makes the CLI target name diverge from the Elixir binding atom :oracle_sql.

Two ways to reconcile:

• Rename the variant OracleSql → Oracle (matches the existing Postgres, Snowflake, Redshift convention; CLI target becomes sql.oracle, Elixir atom becomes :oracle). Updates needed in dialect.rs, cli/test.rs, tests/integration/sql.rs, and the Elixir binding's atom + match arm.
• Keep OracleSql and add #[strum(serialize = "oracle_sql")] on the variant so the user-facing name stays oracle_sql; the snapshot at cli/test.rs:89 then stays correct.

A couple of items missing outside the diff — happy to push a commit for these once the naming question is settled:

• web/book/src/project/target.md lists supported/unsupported dialects; Oracle isn't in either list.
• prqlc/bindings/elixir/lib/prql.ex:42-43 docstring enumerates accepted atoms; :oracle_sql is missing there too.

[prql-bot]

The rename to Dialect::Oracle is missed in the integration test, breaking the build (test-rust (wasm32...) is already failing on this commit). One inline suggestion below.

[max-sixty]

@prql-bot can you add a short note to the docs etc re having a very early version of Oracle dialect?

[prql-bot]

Added a short note in bef16d4 — the dialects list in web/book/src/project/target.md now calls out that sql.oracle is very early (only forces identifier quoting; most other features fall back to generic SQL), and there's a matching CHANGELOG.md entry under Features.