fix: mixed-case validation in hrp::parse_display

[alexgrad42]

Move has_lower and has_upper flags from local variables inside write_str to fields on the ByteFormatter struct. This ensures that case validation persists across multiple write_str calls (e.g., when processing split Display arguments) instead of resetting on each boundary.

Fixes: #276

[tcharding]

Needs unit test. Done as a separate patch last so review can verify the bug this fixes.

[alexgrad42]

I've extracted the test into separate commit, please check

[tcharding]

Thanks, still a couple things to fix please:

• The commit brief log (the title) should not be the same for both commits
• The commit that adds the tests has to be second so that test pass on all commits. The during review someone will re-order the commits and check that the test fails without the fix.

Thanks

[tcharding]

I'm reasonably sure this was mentioned to you on another PR. I'm also still uncertain how much of you work is human and how much LLM. If you are using an LLM you will need to configure it to remember stuff we mention in review otherwise it waists review time. Of course simply forgetting things if you are actually human is ok.

[tcharding]

On further thought the tests are probably too verbose. Why not just use the one in the issue (witthout all the code comments).

[alexgrad42]

I have rearranged the commit and cleaned the test. Note that example from Loupe in fact doesn't work, I've added a case to show this.

[apoelstra]

ACK a5312e7; successfully ran local tests