Improve handling of unknown elements

[coryan]

When I implemented the docfx tool I skipped handling of most elements in a <para>. There are about 300 of them:

https://github.com/doxygen/doxygen/blob/2717489cb90ffd338217b6ce6d2fcb38f25495e6/src/htmlentity.cpp#L40

And we only use about a dozen:

google-cloud-cpp/docfx/doxygen2markdown.cc

Lines 435 to 456 in e6140c8

	bool AppendIfDocTitleCmdGroup(std::ostream& os, MarkdownContext const& ctx,
	pugi::xml_node node) {
	if (AppendIfPlainText(os, ctx, node)) return true;
	if (AppendIfULink(os, ctx, node)) return true;
	if (AppendIfBold(os, ctx, node)) return true;
	// Unexpected: s
	if (AppendIfStrike(os, ctx, node)) return true;
	// Unexpected: underline
	if (AppendIfEmphasis(os, ctx, node)) return true;
	if (AppendIfComputerOutput(os, ctx, node)) return true;
	// Unexpected: subscript, superscript, center, small, del, ins
	// Unexpected: htmlonly, manonly, rtfonly, latexonly, docbookonly
	// Unexpected: image, dot, msc, plantuml
	if (AppendIfAnchor(os, ctx, node)) return true;
	// Unexpected: formula
	if (AppendIfRef(os, ctx, node)) return true;
	// Unexpected: emoji
	if (AppendIfLinebreak(os, ctx, node)) return true;
	// Unexpected: nonbreakablespace
	// Unexpected: many many symbols
	if (AppendIfNDash(os, ctx, node)) return true;
	return false;

The docfx tool intentionally crashes when it finds an element it does not know how to handle. That way we can update the code when needed. But maybe we should proactively add elements that we think may appear in the future and are safe to handle.