Fix HTML escaping after literals in non-code Javadoc (#1409)

When formatting Javadocs, we use a state machine to track what we should
be escaping and when. In the case of non-code blocks, we generally
HTML-escape all characters except `@`, which must be encoded using a
Javadoc `{@literal @}` block. If we see a sequence of `@` characters, we
enter a state where we try to print as many `@`s as possible before
closing the `@literal` block. Unfortunately, there is a bug whereby when
closing the literal block (because we have seen a non-`@` character), we
do not HTML-escape the first character after it. This commit fixes this
issue.

Fixes #1408

CHANGELOG_PENDING.md

@@ -1,4 +1,5 @@
 ### Improvements
 
+### Bug Fixes
 
-### Bug Fixes
+- Fix HTML escaping after `{@literal ...}` blocks in Javadocs

pkg/codegen/java/utilities.go

@@ -148,7 +148,7 @@ func (w *plainJavadocWriter) WriteRune(r rune) {
 			w.b.WriteRune('@')
 		} else {
 			w.b.WriteRune('}')
-			w.b.WriteRune(r)
+			w.b.WriteString(html.EscapeString(string(r)))
 			w.state = plainJavadocText
 		}
 	}

pkg/codegen/java/utilities_test.go

@@ -96,10 +96,12 @@ func TestFormatForeignComments(t *testing.T) {
 - @Foo @ bar
 - Foo @ bar
 - Foo bar @@
+- Foo @<this> works?
 `,
 			expected: autogold.Expect(` * - {@literal @}Foo {@literal @} bar
  * - Foo {@literal @} bar
  * - Foo bar {@literal @@}
+ * - Foo {@literal @}&lt;this&gt; works?
  * `),
 		},
 		{