The short answer is that there is not a universally recognized set of terms to cover your needs within linguistics.
Non-word or nonce are often used in the context of psycholinguistic experiments but they generally refer to word-like strings like "trulky" and measure speed of recognition. They would not apply it to a random string like "hsuwoqjlxs".
The term "token" is also no good because it just refers to string instance as opposed to type. It still mostly implies a word.
As recommended, I'd stick with "string". You can divide them between lexical and non-lexical (meaning they are or are not in your lexicon). You could use pseudo-lexical for nonwords that are not in the lexicon you're comparing your strings to but look like possible English words.