Refactor attribute parsing to borrow UTF-16LE

[omerbenamram]

Summary

• Introduce Utf16LeStr<'a> (UTF-16LE) backed by utf16-simd, and delay UTF-16LE → UTF-8 conversion until display/serialization.
• Refactor MFT attribute parsing to be slice-based and zero-copy where possible.

Notes / Breaking changes

• Several attribute structures are now <'a> and borrow from the entry buffer.

Test plan

• cargo fmt --all -- --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test --all-targets --all-features

---

Note

Introduces borrowed UTF-16LE strings and migrates MFT attribute parsing to slice-based, zero-copy APIs to reduce allocations and improve safety.

• Add utf16-simd and new Utf16LeStr<'a>; delay UTF-16→UTF-8 conversion to display/serialization
• Refactor attribute types to be borrowing (<'a>), e.g. MftAttributeHeader<'a>, FileNameAttr<'a>, AttributeListAttr<'a>, DataAttr<'a>, IndexRootAttr<'a>, RawAttribute<'a>
• Replace stream parsers with from_slice/from_slice_at; implement MftAttributeContent::from_record and use decode_data_runs directly
• Update iteration over attributes to operate on record slices, correctly stop at $END, and add overflow/EOF checks
• Replace consuming into_* helpers with non-consuming as_* accessors; adjust mft_dump, csv, and path building to use borrowed names and to_utf8_string() when needed
• Make timestamp serializers and windows_filetime_to_timestamp public; minor JSON writer callsite cleanups
• Add/extend tests for index root parsing, empty/nonresident mapping pairs, 32-bit overflow, $STANDARD_INFORMATION 48/72-byte layouts, and end-marker handling

^{Written by Cursor Bugbot for commit b38da6e. This will update automatically on new commits. Configure here.}