wikipedia-kyoto-japanese-english: increase REXML entity expansion lim…

…it during XML parsing (red-data-tools#198)

Using `Datasets::WikipediaKyotoJapaneseEnglish#each` raised an `entity
expansion has grown too large (RuntimeError)`. This error occurs because
the entity expansion limit in REXML is set by
ruby/rexml#187, and
`Datasets::WikipediaKyotoJapaneseEnglish#each` exceeds that limit.

In Red Datasets, increasing the entity expansion limit is not a problem
because we want to handle large datasets.
Therefore, we temporarily increase the limit.

## How to reproduce

```console
$ cd red-datasets && bundle
$ bundle exec ruby example/wikipedia-kyoto-japanese-english.rb
...
/home/otegami/.rbenv/versions/3.3.3/lib/ruby/gems/3.3.0/gems/rexml-3.3.4/lib/rexml/parsers/baseparser.rb:560:in `block in unnormalize': entity expansion has grown too large (RuntimeError)
...
```

lib/datasets/wikipedia-kyoto-japanese-english.rb

@@ -89,7 +89,9 @@ def each(&block)
             next unless base_name.end_with?(".xml")
             listener = ArticleListener.new(block)
             parser = REXML::Parsers::StreamParser.new(entry.read, listener)
-            parser.parse
+            with_increased_entity_expansion_text_limit do
+              parser.parse
+            end
           when :lexicon
             next unless base_name == "kyoto_lexicon.csv"
             is_header = true
@@ -106,6 +108,9 @@ def each(&block)
     end
 
     private
+
+    ENTITY_EXPANSION_TEXT_LIMIT = 163_840
+
     def download_tar_gz
       base_name = "wiki_corpus_2.01.tar.gz"
       data_path = cache_dir_path + base_name
@@ -114,6 +119,14 @@ def download_tar_gz
       data_path
     end
 
+    def with_increased_entity_expansion_text_limit
+      default_limit = REXML::Security.entity_expansion_text_limit
+      REXML::Security.entity_expansion_text_limit = ENTITY_EXPANSION_TEXT_LIMIT
+      yield
+    ensure
+      REXML::Security.entity_expansion_text_limit = default_limit
+    end
+
     class ArticleListener
       include REXML::StreamListener