feat(parser): add bytes literal syntax

[thevilledev]

Motivation

Implements bytes literals, which return []byte directly and interpret \x and octal escapes as raw byte values (0-255) rather than Unicode codepoints.

Fixes #457.

Changes

Components:

• lexer/token: add Bytes token kind
• lexer/state: detect [bB]'...' and [bB]"..." prefix, emit Bytes token
• lexer/utils: add unescapeBytes and unescapeByteChar functions for byte-level escape handling (no \u/\U support, 0-255 limit)
• parser: parse Bytes token into BytesNode
• ast/node: add BytesNode struct with []byte value
• ast/visitor: handle BytesNode in AST walker
• ast/print: add String() method
• checker: type inference returns []byte
• compiler: emit push for BytesNode value

Updated docs.

Further comments

This is not exactly CEL-compatible:

• CEL allows \uXXXX in bytes literals (only \U is string-only). We reject both, requiring \x for arbitrary byte values instead.
• CEL supports b\... and triple-quoted bytes. We only support single-line b'...' and b"..."

Non-ASCII characters like b"ÿ" are UTF-8 encoded to []byte{0xC3, 0xBF}, which does match CEL behavior.

Open to discussion whether we want to go for this compatibility or stay at this level!