JavaScript regular expression literals

There is still one broken case that is _very_ hard to fix:

    var re = x / y // comment

Here the string "/ y /" is recognized as a literal while it is not. This
means the comment is not tokenized as a comment, and if it contains any
opening tokens the next line will be incorrectly indented. A working example
is provided in the unittests.

djhtml/modes.py

@@ -366,13 +366,13 @@ class DjJS(DjTXT):
     """
 
     RAW_TOKENS = DjTXT.RAW_TOKENS + [
-        r"\\.",
         r"//.*",
         r"/\*",
         r"[$\w-]+:",
         r'"(?:\\.|[^\\"])*"',  # "string"
         r"'(?:\\.|[^\\'])*'",  # 'string'
         r"`(?:\\.|[^\\`])*`",  # `string`
+        r"/(?:\\.|[^\\/\n])*/",  # /string/
         r"[{[()\]}]",
         r"var ",
         r"let ",

tests/suite/js.html

@@ -436,3 +436,20 @@
             baz
     }
 </script>
+
+
+<!-- Expression literals -->
+<script>
+    re = /ab+c/
+    space_re = / foo /
+    single_re = /foo\/ + bar/
+    double_re = /foo/ + /bar/
+    look_like_comment = /\(//
+
+    // The following is known to be broken
+    const x = 42
+    const y = 7
+    const six = x / (function() { // comment
+    return y
+    })();
+</script>