build: pin file.encoding so the Maven build works on Windows (#2565)

* build: pin file.encoding so the Maven build generates the parser on Windows

JJTree reads the .jjt under GRAMMAR_ENCODING but writes the intermediate .jj
with the JVM default charset. On a platform whose default is a legacy code
page the grammar's Unicode character classes are written lossily, and JavaCC
then rejects the mangled ranges:

    Error: Line 2331, Column 294: Right end of character range '?' has a
    lower ordinal value than the left end of character range '?'.

The Gradle build has pinned -Dfile.encoding=UTF-8 in gradle.properties since
2021, which is why only the Maven build is affected. Give it the same pin via
.mvn/jvm.config and put windows-latest back in the maven_verify matrix.

* build: pin file.encoding for the forked test JVMs as well

ParserKeywordsUtilsTest regenerates the .jj in-process and reads it back, so it
depends on the default charset the same way the build does. Surefire forks its
test JVMs and those do not inherit .mvn/jvm.config, leaving them on the platform
default:

    ParserKeywordsUtilsTest.getAllKeywordsUsingJavaCC ... <<< ERROR!
    org.javacc.parser.MetaParseException

Gradle needs no equivalent: its test workers take their encoding from the daemon,
which gradle.properties already pins.

Author: knutwannheden

.github/workflows/ci.yml

@@ -64,8 +64,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        # windows-latest disabled: see #<issue> — record the reason, not just the comment
-        os: [ ubuntu-latest, macos-latest ]
+        os: [ ubuntu-latest, windows-latest, macos-latest ]
     steps:
       - uses: actions/checkout@v5
         with:

.mvn/jvm.config

@@ -0,0 +1 @@
+-Dfile.encoding=UTF-8

pom.xml

@@ -426,6 +426,11 @@
                         <goals>
                             <goal>jjtree-javacc</goal>
                         </goals>
+                        <!--
+                            GRAMMAR_ENCODING governs reading only: JJTree writes the intermediate
+                            .jj with the JVM default charset, which mangles the grammar's non-ASCII
+                            character classes unless file.encoding is UTF-8 (see .mvn/jvm.config).
+                        -->
                         <configuration>
                             <javaccCmdLineArgs>
                                 <arg>-CODE_GENERATOR:"Java"</arg>
@@ -564,11 +569,13 @@
                         @{jacocoArgLine} is resolved late, after jacoco:prepare-agent has run.
                         Without it the JaCoCo agent would be dropped and coverage reports empty.
                     -->
+                    <!-- Forked test JVMs do not inherit .mvn/jvm.config, so pin the encoding here too. -->
                     <argLine>
                         @{jacocoArgLine}
                         --add-opens=java.base/java.lang=ALL-UNNAMED
                         --add-opens=java.base/java.util=ALL-UNNAMED
                         -Xmx2G -Xms800m -Xss4m
+                        -Dfile.encoding=UTF-8
                     </argLine>
                 </configuration>
             </plugin>