[css-text] Update line-break-{normal,strict}-015.xht for ICU 66 line breaking rules

[litherum]

ICU 66 says that, in ja@lb=normal and ja@lb=strict mode, there are no line break opportunities between the following code points:

• U+30EB and U+3002
• U+3002 and U+2025
• U+2025 and U+2025

[kojiishi]

lgtm, thank you!

[frivoal]

@litherum I'm wondering about these changes. The resulting layout after your changes seems good, but I am not sure it is a "must" level, and if it's not how to deal with testing. The absence of a break between the pairs of U+2025 that the test was originally testing is mandatory (via css-text-3 having a must statement about it), but the other two, while likely good ideas, aren't quite as strongly supported by the spec.

• The absence of a break between U+30EB and U+3002 is listed by UAX-14, but as a tailorable one (i.e. a should). On top of that, css-text-3 says “UA and should follow language conventions”, and disallowing a break between characters with ID class like U+30EB and U+3002 certainly is a strongly established convention. Yet I can find no “must” behind it, so it feels like a strong “should”.

• The absence of a break between U+3002 and U+2025 isn't listed by either css-text-3 or UAX-14. Both css-text-3 and UAX-14 allow for customizations, and this is probably a good one (at least in strict mode), but I can find no “must” to back it up. It feels like it falls somewhere between “may” and “should”.

So, I am wondering if we should replace the current test with two different ones:

• a “should” level test, which is otherwise identical to what we now have after your change
• a “must” level test, Using サンプルサンプル文 instead of サンプルサンプル。 (same length, ends in a ID class character rather than a CL class like U+3002), expecting the break to be after 文 (and thus before the pair of U+2025).

This is not just a theoretical exercise: all browsers used to pass the test before this change, now they all fail it, so we lost a regression test on a must statement for the sake of getting a nudge on a should statement. Presumably, UAs that use ICU will soon recover, but not all do.

[litherum]

That would be fine with me.

[frivoal]

@litherum filed #22458 to do what was suggested above. Review appreciated.