Make double-quoted literal redaction dialect-aware

[coderabbitai]

Requested by @KARTIKrocks as a follow-up to PR #61.

Problem

analyzer.Redact treats every double-quoted run as an identifier. This preserves quoted identifiers in PostgreSQL and MySQL with ANSI_QUOTES, but MySQL's default sql_mode accepts double-quoted strings as literals. As a result, a value such as "alice@example.com" can remain in redacted Result.Query and Result.Fingerprint. This gap predates PR #61 and is documented in website/docs/redaction.md. Changing Redact to redact every double-quoted run would remove useful structure from PostgreSQL ORM-generated queries.

Proposed work

• Define a dialect hint for Analyzer and pass it through the canonical analyzer.Redact / analyzer.Fingerprint path without adding a second normalizer. Keep parser-pluggable analysis and resolve the configuration once during construction, not per query.
• Redact double-quoted strings when MySQL default-mode semantics apply. Preserve double-quoted identifiers when identifier semantics apply, including PostgreSQL and MySQL ANSI_QUOTES. Preserve backtick identifiers.
• Update the relevant analyzer/redact.go and analyzer/analyzer.go behavior, tests in analyzer/redact_test.go, and website/docs/redaction.md. Consider integration coverage for parsers/mysqlparser.

Acceptance criteria

• Under a MySQL default-mode dialect hint, double-quoted literal contents appear in neither redacted Result.Query nor Result.Fingerprint.
• PostgreSQL ORM-style quoted identifiers remain readable; MySQL ANSI_QUOTES identifiers and backtick identifiers remain intact.
• Document the behavior when the dialect is unknown, including any remaining exposure, and test the chosen default. Do not regress single-quoted, dollar-quoted, or numeric redaction.
• Keep the analyzer dependency-light and avoid additional per-query configuration work.

References: #61
Review discussion: #61 (comment)