using regex instead of soup to parse pdf page margins

[shreyashah115]

cherry-pick from develop, related to #4972

[skjbulcher]

I'm thoroughly confused by this commit. It

• artificially limits values to those ending in mm; (it ignores "page-size", who's value doesn't end in mm;), and
• ignores the metric.

Is that intentional? I see more recent code adding options like "header-spacing" and "orientation", which are all ignored as well since their values don't end in mm;.