Self-Test: Ignore width space after em dash (#654)

AsciiDoctor spits out an em dash when there is an em dash between two
words. This is kind of a pleasant surprise because it makes it clear
that we can break words after the em dash. This change teaches the
`html_diff` utility that we use for finding differences in pages that
this particular zero width space doesn't count as a difference.

Author: nik9000

integtest/Makefile

@@ -12,6 +12,7 @@ check: \
 	style \
 	minimal_expected_files minimal_same_files \
 	includes_expected_files includes_same_files \
+	emdash_expected_files emdash_same_files \
 	beta_expected_files beta_same_files \
 	experimental_expected_files experimental_same_files \
 	missing_include_fails_asciidoc missing_include_fails_asciidoctor \

integtest/emdash.asciidoc

@@ -0,0 +1,5 @@
+= Title
+
+== Chapter
+
+I have an em dash between some--words.

integtest/html_diff

@@ -31,6 +31,10 @@ def normalize_html(html):
     # Remove the zero width space that asciidoctor adds after each horizontal
     # ellipsis. They don't hurt anything but asciidoc doesn't make them
     html = html.replace('\u2026\u200b', '\u2026')
+    # Remove the zero width space that asciidoctor adds after each em dash that
+    # is between words. They are actively nice to have but asciidoc doesn't
+    # make them.
+    html = html.replace('\u2014\u200b', '\u2014')
     # Temporary workaround for known issues
     html = re.sub(
         r'(?m)^\s+<div class="console_widget" data-snippet="[^"]+">'