Percent-encoded bytes conformance requirement

[annevk]

@domenic mentioned in #500 that

Sequences of percent-encoded bytes, string percent decoded, should not cause UTF-8 decode without BOM or fail to return failure.

Is confusing. What it is mainly trying to say is that, e.g., "%FF" should not be used as 0xFF would end up as U+FFFD when UTF-8 decoded (or return failure for the above cited algorithm). There might be edge cases however where encoding non-string data in URLs is fine.

But perhaps we should simply stay out of this?