fix(devin-lang): ensure proper handling of language identifiers #101

In the previous version, the lexer incorrectly switched to the initial state upon encountering an identifier within a language identifier comment block, causing the language identifier to be interpreted as a code block start. This commit corrects the behavior by directly returning the LANGUAGE_ID token when a valid language identifier is found, without switching to the initial state.

docs(docs/scenes/legacy-migration.md): add screenshot and instructions for generating test data

This commit enhances the legacy migration documentation by adding a visual guide and step-by-step instructions for generating test data using the "Generate Test Data (APIs)" feature. The updated documentation includes a screenshot of the AutoDev Living Documentation, illustrating the process of generating API test data from Java Controller code.

docs/scenes/legacy-migration.md

@@ -20,6 +20,8 @@ Building upon the experience accumulated in our "[System Refactoring and Migrati
 
 Support language: Java.
 
+![AutoDev Living Documentation](https://unitmesh.cc/auto-dev/gen-test-data.png)
+
 Simply right-click on the Java Controller code, select "Generate Test Data (APIs)," and API test data will be generated.
 
 ### 2. Custom Prompt Integration

exts/devin-lang/src/grammar/DevInLexer.flex

@@ -103,6 +103,6 @@ NEWLINE= \n | \r | \r\n
 
 <LANG_ID> {
    "```"             { return CODE_BLOCK_START; }
-   {IDENTIFIER}      { yybegin(YYINITIAL); return LANGUAGE_ID;  }
+   {LANGUAGE_ID}     { return LANGUAGE_ID;  }
    [^]               { yypushback(1); yybegin(CODE_BLOCK); }
 }