Introduce StringEscapeOptions, FallbackEscapeFormat, and QuoteEscapeMethod
types that declaratively describe any string literal escape grammar.

Add RegexBuilder::string_escape() method that uses these options to
transform a regex R into R' such that strings matching R' are valid
string literals whose deserialized content matches R.

Rewrite json_quote() as a thin wrapper that converts JsonQuoteOptions
into StringEscapeOptions and delegates to string_escape(). The
JsonQuoteOptions public API is unchanged from main.

New capabilities over the old JSON-only API:
- Configurable single-char escape mappings (e.g. \a, \v, \0)
- Configurable fallback escape format (\uXXXX or \xHH)
- Configurable quote character
- Configurable quote escape method (backslash or doubling)
- Configurable must-escape byte set

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The quote_char field is typed as char but the implementation operates on
of string_escape() to reject non-ASCII quote characters with a clear
error message instead of silently producing wrong behavior.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

In Doubling mode (e.g. YAML single-quoted strings), backslash is not an
escape prefix and should pass through as a literal character. Previously,
backslash was unconditionally added to must_escape_set and emitted as
\\\\ regardless of escape mode.

Now backslash is only treated as special when the grammar actually uses
it as an escape prefix (has single_char_escapes, has fallback format, or
uses Backslash quote escape method).

Add test verifying backslash is literal in Doubling mode.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The escape logic previously only handled bytes 0x00-0x1F in the inner
loop and hardcoded 0x7F separately. Bytes outside this range that
appeared in must_escape would be silently dropped.

Changes:
- Generalize the slow-path loop in quote_byteset to iterate all 256
  bytes in must_escape_set, not just 0..32
- Replace `bs_without_ctrl[0] = 0` (blanket clear of bits 0-31) with
  a targeted clear of only must_escape bytes
- Add a post-loop that handles must_escape bytes >= 0x20 that aren't
  covered by the control-char fast paths (quote_all_ctrl)
- This correctly handles 0x7F and any other non-control must_escape
  bytes via the same generalized path

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The UnicodeXXXX and HexHH branches were identical except for the prefix
string. Extract the prefix into a variable and share the hex digit regex
construction logic.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

This mirrors the existing JsonQuote variant but uses the new
StringEscapeOptions. Users can now build a RegexAst tree containing
StringEscape nodes directly without going through RegexBuilder methods.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Sort single_char_escapes and must_escape before comparing/hashing so
that two options with the same entries in different orders produce the
same cache key.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Avoid re-computing the conversion on every json_quote call by caching
the result in a HashMap keyed by JsonQuoteOptions (now derives Hash/Eq).

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

UnicodeXXXX (\u00XX) can only correctly represent bytes 0x00-0x7F.
For bytes >= 0x80 the Unicode code point differs from the raw byte value,
causing a silent roundtrip mismatch. Now returns an error unless the byte
has an explicit single_char_escape mapping.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Avoids unconditionally flagging backslash as needing escape in Doubling
mode where backslash is a literal character.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Only include qc and backslash in the escape char byteset when they fall
in the control range (< 0x20). Since both are always >= 0x20 for typical
configurations, this removes spurious alternatives from the fast-path
regex that would accept escape sequences for non-control bytes.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

The slow path iterates all 256 bytes so the post-loop for bytes >= 0x20
was producing duplicate regex alternatives. Now only runs after the fast
path which only covers 0x00-0x1F.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Replace custom Hash/PartialEq impls with derived ones and add a
normalize() method that sorts and dedups the Vec fields. Called once
at string_escape entry before cache lookup.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- must_escape byte 0x7F with HexHH (tests >= 0x20 escape path)
- UnicodeXXXX rejection for must_escape bytes >= 0x80
- Control-only regex (tests fast path correctness)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

This reverts commit c7af1873e08b9c053520f4644db755e006e0623e.

Remove has_fallback parameter from quote_byteset (computed from options
internally). Apply rustfmt formatting across changed files.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

- Expand docstrings on StringEscapeOptions, FallbackEscapeFormat,
  QuoteEscapeMethod, JsonQuoteOptions, and the RegexAst variants
- Document backslash handling semantics, normalization, error
  conditions, and the interaction between quote_escape and
  uses_backslash
- Clarify that \\ and quote_char in allowed_escapes are always
  applied during JsonQuoteOptions conversion (not gated)
- Remove dead branch (b'\\' < 0x20 is always false)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Replace single_char_escapes, escape_prefix, quote_char, quote_escape,
FallbackEscapeFormat, and QuoteEscapeMethod with:
- escape_sequences: Vec<(u8, Vec<u8>)> for explicit per-byte escapes
- fallback_prefix: Option<Vec<u8>> + hex digits for remaining bytes
- max_fallback_byte: Option<u8> for validation
- must_escape: Vec<u8> (fully explicit, no implicit additions)

Remove raw_mode (delimiter wrapping moved to json_quote).
Remove FallbackEscapeFormat and QuoteEscapeMethod enums entirely.

Bytes with both an explicit escape and a fallback now accept either
form (e.g., JSON \b and \u0008 both valid for byte 0x08).

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>