Normalize multipart part header names to lower-case bytes before storing them internally, then look up content-disposition and content-transfer-encoding using lower-case keys.

This keeps the parser aligned with MIME multipart behavior and matches the approach used by Django for multipart part headers. It avoids converting header names to strings or title-casing them.

• Closes Treat headers case insenitively, internally #190
• Closes Headers are being treated case sensitive #165