js: Check metadata string limits in UTF-8 bytes

[ara-stock]

Problem

updateTokenMetadataInstruction in clients/js checks the name, symbol and URI limits with tokenName.length > 32, tokenSymbol.length > 10 and tokenUri.length > 200. These count UTF-16 code units, but the Metaplex limits are in bytes (the CLI's is_valid_token_* already uses s.len()). So a non-ASCII value can pass the client check and then fail on chain. For example, 'あ'.repeat(11) is 11 characters but 33 bytes: the client builds the instruction, and the transaction fails with Metaplex custom program error: 0xb (NameTooLong). The same happens for a 12-byte symbol (0xc) and a 201-byte URI (0xd).

Summary of Changes

• Encode the strings with TextEncoder before the checks (the encoded bytes were already used for the instruction data since js: Use UTF-8 byte length for metadata string prefixes #859) and compare the byte lengths against the limits. The error messages now say "bytes".
• Add a js-legacy test that updateTokenMetadata rejects a name, symbol and URI that are within the character count but over the byte limit.

I used an AI assistant (Claude) while preparing this. I reproduced the issue and checked the fix and tests myself.

🤖 Generated with Claude Code