Reimplement MirthPlainDocument without Sun-derived code

[tonygermano]

Fixes #473

org.syntax.jedit.MirthPlainDocument was a 2007 copy of the Java 1.5 javax.swing.text.PlainDocument source, and it still carried Sun's "PROPRIETARY/CONFIDENTIAL" header, which is not compatible with the project's MPL-2.0 license. This PR replaces it with an original implementation that extends PlainDocument instead of copying it.

What the old copy changed

Compared with the JDK 17 and 25 PlainDocument (which are identical apart from one javadoc tag), the copy only changed two things:

• Lines break on a lone CR as well as on LF, with CRLF counted as one break, so CR-delimited HL7 shows one segment per line in the jEdit text area.
• The default tab size is 4 instead of 8.

Everything else was the JDK's code, plus private copies of two package-private helpers it needed because it lived outside javax.swing.text.

How the new class works

• MirthPlainDocument now extends PlainDocument and sets the tab size to 4.
• PlainDocument finds line breaks with a single read of the inserted text, splitting after each LF. The document's content is a small GapContent subclass that reports each lone CR as an LF for that one read only, using a copy of the text. PlainDocument therefore puts the breaks in the same places, the stored text is unchanged, and each insert still produces a single root element change, which JEditTextArea and SyntaxDocument rely on.
• The translation is tied to the inserting thread and is cleared after the first read, so getText, painting, tokenizing, and the bidi and multibyte checks in AbstractDocument always see the real characters.
• The class moved from the third-party org.syntax.jedit package to com.mirth.connect.client.ui.components, next to MirthSyntaxTextArea.
• The unused MirthPlainDocument(Content) constructor was removed.

Commits

1. Add tests for MirthPlainDocument line breaks. These are characterization tests, committed first and passing against the old Sun-derived class. They cover:
   • CR, LF, CRLF and LFCR line breaks
   • joining and splitting CRLF through edits
   • one root element change per insert
   • removing text across lines
   • undo and redo
   • the stored text being unchanged
   • the tab size
   • right-to-left Hebrew and Arabic lines and their bidi levels, and mixed-direction lines
   • bidi control characters
   • CJK and supplementary characters
   • input-method composed text
2. Reimplement MirthPlainDocument on top of PlainDocument. Replaces the class. The same tests pass unchanged.

Known behavior (unchanged)

The old class did both of these too:

• Inserting a lone CR immediately before an existing LF (for example, pasting text that ends in a bare CR at the end of an LF-terminated line) shows the LF on its own line until the text is reloaded.
• Deleting a character between a CR and an LF leaves them on two lines until the text is reloaded.

In both cases the stored text is correct, so only the display is affected.

Testing

• ./gradlew :client:test passes.
• A side-by-side fuzz comparison of the old and new classes ran 100,000 random insert and remove sequences, including composed, bidi and multibyte text. Line maps, element changes and text were identical in every case.
• Manual testing in the client:
  • the transformer's inbound message template
  • the Channel Writer template
  • the Send Message dialog

Tip: If testing manually, right click in the text area and select "Show Line Endings."

🤖 Generated with Claude Code

[github-actions]

Test Results

129 files  + 1  129 suites  +1   4m 0s ⏱️ + 1m 35s
748 tests +19  748 ✅ +19  0 💤 ±0  0 ❌ ±0 
838 runs  +19  832 ✅ +19  6 💤 ±0  0 ❌ ±0 

Results for commit 1264fd0. ± Comparison against base commit a13e6af.