Write multi-character annotation mnemnonics

WFDB annotation mnemonics ("symbols", in wfdb-python parlance) are not limited to a single character. (All of the standard mnemonics are a single character, but a custom mnemonic defined using setannstr can be multiple characters; the only limitation is that it can't contain spaces or control characters.)

Thanks to pull request #404 it is now possible to define custom annotation mnemonics when calling wfdb.wrann. However, custom_triplet_bytes will choke if any of them are not a single character (something that can be passed to ord).

(interpret_defintion_annotations will read multi-character mnemonics correctly, though note that it has the same problem as rdheader used to have: it uses findall when it should be using match or fullmatch.)

Start by locating custom_triplet_bytes and tracing how wfdb.wrann passes custom annotation mnemonics into it. Check the existing handling in interpret_defintion_annotations for multi-character values, then verify that custom mnemonics without spaces or control characters can be written successfully.