test: cover line subscriber UTF-8 and newline boundaries

[crowndlt]

Adds eight tests for UTF-8 decoding and line boundaries in BodySubscribers.fromLineSubscriber, following the discussion in requests-scala#156.

The cases cover multibyte characters split across buffers and onNext calls, byte-at-a-time delivery, CR/LF/CRLF with the default separator, and an incomplete UTF-8 sequence at EOF. An explicit LF separator is included as a control.

Validation on Linux:

• JVM 25: 8 passed.
• Scala Native: 1 passed, 7 failed. The explicit LF control passes; the other cases expose the decoding and default-separator differences.

Keeping this as a draft because it adds reproducing tests and the implementation still needs fixing. The PR changes only the test file.

To run the suite, use modules.java-http-client.jvm.test.testOnly snhttp.test.java.net.http.BodySubscribersLineTest or modules.java-http-client.test.testOnly snhttp.test.java.net.http.BodySubscribersLineTest with Mill.

For any follow-up fixes, would these issues fall within the scope of the original bounty?

[lqhuang]

Close since it seems agent spam. Reopen only if you're not a bot.

[crowndlt]

No worries, I saw the original bounty was withdrawn, so I'll leave it here. Thanks for taking a look!

[lqhuang]

Hi @crowndlt, Sorry, I've reopened this PR. Too many bots suddenly appeared for unknown reasons. I will take a look soon. Thanks for your contribution!