Incorrect title detection on some pages (SVG title instead of HTML title)

[francoishumanoid]

When I share an url on Shaarli, it uses the <title> inside svg instead of HTML Title.

https://www.numerama.com/sciences/1247146-aller-sur-mars-en-45-jours-cest-la-promesse-de-ce-propulseur-nucleaire.html

[nodiscc]

On my instance (Shaarli 0.12.1 installed from release zip),

• Shaaring this page from the bookmarklet or Firefox extension: the title is correctly detected Aller sur Mars en 45 jours, c'est la promesse de ce propulseur nucléaire - Numerama (copied from the tab title)
• Shaaring this page from the + Shaare button -> paste URL -> Add link -> the title is populated as Numerama.

I think the title extraction regex

Shaarli/application/bookmark/LinkUtils.php

Lines 12 to 18 in 820cae2

	function html_extract_title($html)
	{
	if (preg_match('!<title.*?>(.*?)</title>!is', $html, $matches)) {
	return trim(str_replace("\n", '', $matches[1]));
	}
	return false;
	}

indeed matches something it shouldn't (the SVG <title>):

$ curl -s https://www.numerama.com/sciences/1247146-aller-sur-mars-en-45-jours-cest-la-promesse-de-ce-propulseur-nucleaire.html|grep '<title>'
	<title>Aller sur Mars en 45 jours, c'est la promesse de ce propulseur nucléaire - Numerama</title>
        <symbol id="logo-full" viewbox="0 0 509 49" xmlns="http://www.w3.org/2000/svg"><title>Numerama</title><path d="M240.55 48.462c6.658 0 13.499-1.651...
        <title>Numerama, le média de référence sur la société numérique et l'innovation technologique</title>

return trim(str_replace("\n", '', $matches[1])); should return the first match for <title>, but in this case it returns the second.