Add 63 XML spec edge-case tests + 7 fuzz targets to xmlt
Parser edge cases (from W3C XML 1.0 spec):
- Character references (hex, decimal, non-ASCII: €, ☃)
- Entity references in text and attributes
- Attribute whitespace normalization, quoting variants
- CDATA with special chars, split ]]> escape
- Deep nesting (200 levels), large content (1MB), many attrs (200)
- Mixed content, processing instructions, DOCTYPE variants
- 14 error cases: unclosed/mismatched tags, bare <&, null bytes
Codec edge cases:
- Missing required attrs/children → Error
- Extra unknown attrs/children → Ok (ignored)
- Int/float/bool type errors, NaN/inf handling
- Nested records, special characters in roundtrips
Fuzz enhancements:
- Random valid XML generation with custom generators
- Codec decode crash safety for all primitive types
- Codec roundtrip with random values
131 tests + 9 fuzz targets, all pass.