feat: architect zero-allocation XDR stream parser

[kosisochukwu1234]

Closes #416

Replaces DOM-based XDR parsing with a zero-allocation, cursor-based byte reader (xdr_parser.rs). This parser extracts relevant fields (like transaction IDs) sequentially from raw XDR streams without allocating intermediate heap structures for the entire payload.

Architectural Changes

• Cursor Implementation: Introduced Cursor<'a> in xdr_parser.rs to wrap &[u8], providing strictly bounds-checked methods for reading u32, i32, bool, u64, and opaque byte slices.
• Zero-Allocation: The parser operates entirely in #![no_std] style. It avoids Vec or String allocations by reading directly from the underlying byte slice and extracting sub-slices for variable-length opaque data.
• Skipping Discarded Fields: Added a skip method to efficiently bypass irrelevant union arms and unneeded fields, ensuring discarded data doesn't consume memory or CPU cycles.

Safety & Constraints

• Strict Bounds Checking: Every read operation verifies remaining() bytes against the requested length, preventing out-of-bounds panics. Returns XdrError::UnexpectedEndOfStream on failure.
• Padding Validation: XDR padding boundaries are mathematically calculated and explicitly verified to be zero, returning XdrError::InvalidPadding for malformed data.
• Graceful Error Handling: Rejects malformed or incomplete streams via the XdrError enum instead of unwrapping or panicking.

Module Integration

• Exported the new xdr_parser module in packages/contracts/src/lib.rs.
• Added comprehensive unit tests for endianness conversion, boolean bounds (strictly 0 or 1), length prefixing, padding alignment, and stream exhaustion.

[Damola-Sodiq]

Reviewed