fix(shim): propagate exit code instead of reporting failure on windows

On Windows, syscall.Exec is not supported so the shim falls back to
cmd.Run(). When the wrapped command exits with a non-zero code,
cmd.Run() returns an *exec.ExitError which was incorrectly treated
as a shim failure.

Now the shim detects ExitError and calls os.Exit() with the command's
actual exit code instead of printing a misleading error message.

Fixes #59

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: CalvinAllen

CLAUDE.md

@@ -0,0 +1,13 @@
+# Claude Code Instructions for dtvem
+
+## Deployment
+
+After building, always deploy both executables to the user's installation directory:
+
+```bash
+cp dist/dtvem.exe ~/.dtvem/bin/dtvem.exe
+cp dist/shim.exe ~/.dtvem/bin/dtvem-shim.exe
+~/.dtvem/bin/dtvem.exe reshim
+```
+
+The `reshim` command must be run after deploying the shim to regenerate all runtime shims (npm, node, python, etc.) with the updated binary.

src/cmd/shim/main.go

@@ -311,7 +311,17 @@ func executeCommand(execPath string, args []string) error {
 			Stdout: os.Stdout,
 			Stderr: os.Stderr,
 		}
-		return cmd.Run()
+		if err := cmd.Run(); err != nil {
+			// Check if this is an exit error (command ran but returned non-zero)
+			var exitErr *exec.ExitError
+			if errors.As(err, &exitErr) {
+				// Command executed successfully but returned non-zero exit code
+				// This is not a shim error - propagate the exit code
+				os.Exit(exitErr.ExitCode())
+			}
+			// Other errors (couldn't start, etc.) are actual failures
+			return err
+		}
 	}
 
 	return nil