Some Recorder-generated text asserts need improved unicode normalization

[mdmintz]

Some Recorder-generated text asserts need improved unicode normalization.
(https://unicode.org/reports/tr15/#Norm_Forms has details for anyone reading this that's confused.)

Essentially, some foreign language characters can be represented by more than one unicode format, and the Recorder isn't being consistent with character generation, which may cause text-based asserts to fail after the Recorder generates self.assert_text(TEXT, SELECTOR) lines because even if the TEXT appears to match the visible text on the web page, different unicode formats could cause the assertion to fail.

Here's an example of that:

ipdb> 'й'.encode()
b'\xd0\xb9'
ipdb> 'й'.encode()
b'\xd0\xb8\xcc\x86'